Support
Rankings

Scoring methods

A mix is a set of benchmarks with weights. A scoring method turns a model’s results on that mix into one number. Priors has three. The weighted average is the default and what every ranking uses unless you change it under Calculation settings in the mix editor. A ranking that uses another method says so beside its scores.

All three rank the same models. Whether a model has enough measured data to be ranked is decided first, by the same rules, so switching method never adds or removes a model.

Weighted average

Each benchmark’s score, on a 0 to 100 scale, times its weight, added up. Percent scores are used as published. A benchmark at 15% contributes 15% of the score.

Fable 5.2 max at 64.5, benchmark by benchmarkA track is as long as the benchmark's weight and filled to the model's score. The filled lengths add up to the index score.
Terminal-Bench-Science-AA73.3 × 15%11.00
AA-Briefcase v1.164.2 × 15%9.62
HLE-Diamond80.2 × 10%8.02
AA-Omniscience v1.1 Accuracy79.1 × 10%7.91
CritPt v2 Hard Subset76.5 × 10%7.65
FrontierCode 2.0-AA48.8 × 15%7.31
Terminal-Bench v4-AA59.6 × 10%5.96
AutomationBench-AA76.4 × 5%3.82
AA-Omniscience v1.1 Non-hallucination40.7 × 5%2.04
GDP.pdf24.2 × 5%1.21
Added up64.5

When a model has no result on a benchmark, the gap is filled with an estimate from similar models, so a model is not rewarded for skipping its weakest benchmarks.

Limits. It treats a point on an easy benchmark and a point on a hard one as the same thing. Going from 90 to 95 on a nearly solved benchmark counts as much as going from 10 to 15 on one almost nobody can do. And the filled gaps are guesses that look as exact as measurements.

The same five points, three sizes of stepScores on a benchmark follow an S-shaped curve as models get more able. Five points in the middle of the curve take the least ability. Five points near the floor take 2.3 times as much, and five near the ceiling 3.7 times. The weighted average counts all three the same.
Score10 to 152.3×50 to 551.0×90 to 953.7×Ability each step takes, against the step in the middle
Score10 to 152.3×50 to 551.0×90 to 953.7×Ability each step takes

IRT ability

Item response theory comes from exam scoring. It assumes each model has one underlying ability and each benchmark has a difficulty and a steepness: how much ability it takes to score halfway, and how quickly scores rise past that point. Priors works out those two numbers for every benchmark from all the models that took it, then finds the ability that best explains a model’s results on your mix.

Each benchmark gets a curveThe General frontier mix, fitted on 2026-11-22 to 82 models. A dot is one model's score on the benchmark of that colour, placed at the model's ability. Where a curve crosses halfway is the benchmark's difficulty, and how fast it climbs is its steepness.
AutomationBench-AA halfway at 0.08Terminal-Bench-Science-AA halfway at 0.54GDP.pdf halfway at 1.20the mix’s other 7
0255075100-2-1.5-1-0.500.511.5Score on the benchmarkAbility (0 is the middle of the models Priors tracks)Top model todayGemini 4 Neon xhigh: 79.3 on AutomationBench-AAFable 5.2 max: 76.4 on AutomationBench-AAGemini 4 Carbon high: 77.8 on AutomationBench-AAGPT-6.2 Astra max: 77.2 on AutomationBench-AAOpus 5.5 max: 69.5 on AutomationBench-AAGemini 4 Hydrogen xhigh: 76.1 on AutomationBench-AAGPT-6 Astra max: 68.5 on AutomationBench-AAGPT-6.2 Sol max: 71.3 on AutomationBench-AASonnet 5.5 max: 71.8 on AutomationBench-AAGPT-6.1 Sol max: 64.9 on AutomationBench-AAGemini 4 Argon high: 77.5 on AutomationBench-AAMuse Wonder Full max: 71.3 on AutomationBench-AAGPT-6.1 Terra max: 60.2 on AutomationBench-AAFable 5.1 xhigh: 57.8 on AutomationBench-AAKimi K3.1 max: 73.3 on AutomationBench-AAMuse Wonder max: 68.1 on AutomationBench-AAQwen 4 Max xhigh: 72.6 on AutomationBench-AAOpus 5 max: 56.6 on AutomationBench-AAGPT-6 Sol max: 61.6 on AutomationBench-AAGPT-5.6 Sol max: 60.1 on AutomationBench-AAMuse Ember max: 68.1 on AutomationBench-AAMuse Spark 1.3 max: 57.9 on AutomationBench-AAGrok 4.7 high: 63.5 on AutomationBench-AAHaiku 5.5 max: 35.4 on AutomationBench-AAQwen 3.8 Max xhigh: 56.2 on AutomationBench-AAInkling 2 Small: 45.6 on AutomationBench-AAGrok 4.6 xhigh: 67.0 on AutomationBench-AAGLM 5.3 max: 62.2 on AutomationBench-AADeepSeek V4.1 Flash 1017 max: 66.6 on AutomationBench-AAHy4: 65.6 on AutomationBench-AALing 3.1 Flash: 67.5 on AutomationBench-AAGemini 3.8 Flash high: 59.9 on AutomationBench-AAStep 5 Preview: 51.0 on AutomationBench-AADeepSeek V4.1 Flash max: 68.9 on AutomationBench-AAQwen 3.8 2.4T-A95B: 57.2 on AutomationBench-AAMiMo V2.6 Pro: 58.6 on AutomationBench-AAQwen 3.8 Flash: 55.9 on AutomationBench-AAKimi K3 max: 58.3 on AutomationBench-AAGLM 5.3 Flash max: 60.4 on AutomationBench-AAGPT-5.6 Terra max: 59.6 on AutomationBench-AAMistral Large 4 high: 57.8 on AutomationBench-AAMuse Spark Open xhigh: 49.7 on AutomationBench-AAJoyAI-VL-Preview max: 61.7 on AutomationBench-AAMiMo V2.6 Flash: 64.1 on AutomationBench-AASonnet 5 max: 36.5 on AutomationBench-AAGPT-6 Luna max: 53.2 on AutomationBench-AAMiniMax M3.1 max: 72.5 on AutomationBench-AAAgnes 2.5 Pro Theta: 48.8 on AutomationBench-AAGLM 5.2 max: 28.4 on AutomationBench-AANanbeige4.3-33B: 53.1 on AutomationBench-AAQwen 3.8 27B xhigh: 48.2 on AutomationBench-AAInkling 2: 41.8 on AutomationBench-AAGPT-5.6 Luna max: 50.2 on AutomationBench-AAK2 Horizon 375B A23B high: 37.2 on AutomationBench-AADeepSeek V4 Pro 0813 max: 56.7 on AutomationBench-AABeam 1.0: 48.5 on AutomationBench-AAJT-4.1 Flash 236B A21B: 36.7 on AutomationBench-AASolar Pro 4: 13.6 on AutomationBench-AAMiniMax M3: 21.3 on AutomationBench-AAK2 Horizon MoVA 36B A4B: 26.1 on AutomationBench-AANemotron 3.5 Super VL high: 14.8 on AutomationBench-AAK2 Horizon 32B: 29.4 on AutomationBench-AANemotron 3 Ultra max: 3.0 on AutomationBench-AASolar Mini 4: 22.3 on AutomationBench-AAGemini 3.5 Flash Lite high: 25.0 on AutomationBench-AAInkling Small: 4.8 on AutomationBench-AAApodex 1.1: 28.2 on AutomationBench-AAK2 Horizon 7B: 20.9 on AutomationBench-AAInkling xhigh: 5.0 on AutomationBench-AAA.X-K2: 2.5 on AutomationBench-AAK-EXAONE 2.0: 3.9 on AutomationBench-AALFM3-1.7B: 7.0 on AutomationBench-AAK2 Horizon 3.7B: 14.7 on AutomationBench-AAGranite 4.2 30B: 1.5 on AutomationBench-AANemotron 3.5 Lightning: 0.8 on AutomationBench-AAGranite 4.2 8B: 2.1 on AutomationBench-AAGranite 4.2 3B: 0.4 on AutomationBench-AAMuse Glimmer high: 6.8 on AutomationBench-AALing 3.0 Tiny: 0.3 on AutomationBench-AANorth Mini Code: 1.5 on AutomationBench-AAK2 Horizon 0.9B: 0.2 on AutomationBench-AALFM3-VL-2B: 0.2 on AutomationBench-AAGemini 4 Neon xhigh: 70.0 on Terminal-Bench-Science-AAFable 5.2 max: 73.3 on Terminal-Bench-Science-AAGemini 4 Carbon high: 64.8 on Terminal-Bench-Science-AAGPT-6.2 Astra max: 78.1 on Terminal-Bench-Science-AAOpus 5.5 max: 59.0 on Terminal-Bench-Science-AAGemini 4 Hydrogen xhigh: 49.5 on Terminal-Bench-Science-AAGPT-6 Astra max: 63.3 on Terminal-Bench-Science-AAGPT-6.2 Sol max: 67.6 on Terminal-Bench-Science-AASonnet 5.5 max: 53.3 on Terminal-Bench-Science-AAGPT-6.1 Sol max: 58.1 on Terminal-Bench-Science-AAGemini 4 Argon high: 52.9 on Terminal-Bench-Science-AAMuse Wonder Full max: 37.1 on Terminal-Bench-Science-AAGPT-6.1 Terra max: 57.6 on Terminal-Bench-Science-AAFable 5.1 xhigh: 40.5 on Terminal-Bench-Science-AAKimi K3.1 max: 34.3 on Terminal-Bench-Science-AAMuse Wonder max: 25.7 on Terminal-Bench-Science-AAQwen 4 Max xhigh: 23.3 on Terminal-Bench-Science-AAOpus 5 max: 28.6 on Terminal-Bench-Science-AAGPT-6 Sol max: 30.0 on Terminal-Bench-Science-AAGPT-5.6 Sol max: 22.4 on Terminal-Bench-Science-AAMuse Ember max: 20.0 on Terminal-Bench-Science-AAMuse Spark 1.3 max: 11.0 on Terminal-Bench-Science-AAGrok 4.7 high: 9.5 on Terminal-Bench-Science-AAHaiku 5.5 max: 20.0 on Terminal-Bench-Science-AAQwen 3.8 Max xhigh: 11.9 on Terminal-Bench-Science-AAInkling 2 Small: 5.7 on Terminal-Bench-Science-AAGrok 4.6 xhigh: 2.9 on Terminal-Bench-Science-AAGLM 5.3 max: 9.5 on Terminal-Bench-Science-AADeepSeek V4.1 Flash 1017 max: 12.9 on Terminal-Bench-Science-AAHy4: 12.9 on Terminal-Bench-Science-AALing 3.1 Flash: 4.3 on Terminal-Bench-Science-AAGemini 3.8 Flash high: 10.0 on Terminal-Bench-Science-AAStep 5 Preview: 2.4 on Terminal-Bench-Science-AADeepSeek V4.1 Flash max: 9.0 on Terminal-Bench-Science-AAMiMo V2.6 Pro: 5.7 on Terminal-Bench-Science-AAQwen 3.8 Flash: 7.1 on Terminal-Bench-Science-AAKimi K3 max: 8.1 on Terminal-Bench-Science-AAGLM 5.3 Flash max: 4.8 on Terminal-Bench-Science-AAGPT-5.6 Terra max: 13.8 on Terminal-Bench-Science-AAMistral Large 4 high: 4.3 on Terminal-Bench-Science-AAMuse Spark Open xhigh: 1.9 on Terminal-Bench-Science-AAJoyAI-VL-Preview max: 3.8 on Terminal-Bench-Science-AAMiMo V2.6 Flash: 4.3 on Terminal-Bench-Science-AAGPT-6 Luna max: 8.6 on Terminal-Bench-Science-AAMiniMax M3.1 max: 3.3 on Terminal-Bench-Science-AAAgnes 2.5 Pro Theta: 3.8 on Terminal-Bench-Science-AANanbeige4.3-33B: 1.0 on Terminal-Bench-Science-AAQwen 3.8 27B xhigh: 0.5 on Terminal-Bench-Science-AAInkling 2: 1.4 on Terminal-Bench-Science-AAGPT-5.6 Luna max: 3.3 on Terminal-Bench-Science-AAK2 Horizon 375B A23B high: 0.0 on Terminal-Bench-Science-AADeepSeek V4 Pro 0813 max: 5.7 on Terminal-Bench-Science-AABeam 1.0: 0.5 on Terminal-Bench-Science-AAJT-4.1 Flash 236B A21B: 0.0 on Terminal-Bench-Science-AASolar Pro 4: 0.0 on Terminal-Bench-Science-AAMiniMax M3: 0.5 on Terminal-Bench-Science-AAK2 Horizon MoVA 36B A4B: 0.0 on Terminal-Bench-Science-AANemotron 3.5 Super VL high: 0.0 on Terminal-Bench-Science-AAK2 Horizon 32B: 0.0 on Terminal-Bench-Science-AANemotron 3 Ultra max: 0.0 on Terminal-Bench-Science-AAGemini 3.5 Flash Lite high: 0.0 on Terminal-Bench-Science-AAInkling Small: 0.0 on Terminal-Bench-Science-AAApodex 1.1: 0.5 on Terminal-Bench-Science-AAK2 Horizon 7B: 0.0 on Terminal-Bench-Science-AAInkling xhigh: 0.0 on Terminal-Bench-Science-AAA.X-K2: 0.0 on Terminal-Bench-Science-AAK-EXAONE 2.0: 0.0 on Terminal-Bench-Science-AALFM3-1.7B: 0.0 on Terminal-Bench-Science-AAK2 Horizon 3.7B: 0.0 on Terminal-Bench-Science-AAGranite 4.2 30B: 0.0 on Terminal-Bench-Science-AANemotron 3.5 Lightning: 0.0 on Terminal-Bench-Science-AAGranite 4.2 8B: 0.0 on Terminal-Bench-Science-AAGranite 4.2 3B: 0.0 on Terminal-Bench-Science-AAMuse Glimmer high: 0.0 on Terminal-Bench-Science-AANorth Mini Code: 0.0 on Terminal-Bench-Science-AAK2 Horizon 0.9B: 0.0 on Terminal-Bench-Science-AALFM3-VL-2B: 0.0 on Terminal-Bench-Science-AAGemini 4 Neon xhigh: 28.0 on GDP.pdfFable 5.2 max: 24.2 on GDP.pdfGemini 4 Carbon high: 28.4 on GDP.pdfGPT-6.2 Astra max: 29.8 on GDP.pdfOpus 5.5 max: 26.2 on GDP.pdfGemini 4 Hydrogen xhigh: 27.6 on GDP.pdfGPT-6 Astra max: 31.0 on GDP.pdfGPT-6.2 Sol max: 30.0 on GDP.pdfSonnet 5.5 max: 25.8 on GDP.pdfGPT-6.1 Sol max: 31.0 on GDP.pdfGemini 4 Argon high: 21.8 on GDP.pdfMuse Wonder Full max: 32.6 on GDP.pdfGPT-6.1 Terra max: 28.6 on GDP.pdfFable 5.1 xhigh: 26.2 on GDP.pdfKimi K3.1 max: 31.4 on GDP.pdfMuse Wonder max: 30.6 on GDP.pdfQwen 4 Max xhigh: 26.2 on GDP.pdfOpus 5 max: 21.6 on GDP.pdfGPT-6 Sol max: 25.2 on GDP.pdfGPT-5.6 Sol max: 27.2 on GDP.pdfMuse Ember max: 22.8 on GDP.pdfMuse Spark 1.3 max: 26.6 on GDP.pdfGrok 4.7 high: 23.2 on GDP.pdfHaiku 5.5 max: 20.8 on GDP.pdfQwen 3.8 Max xhigh: 22.8 on GDP.pdfInkling 2 Small: 25.2 on GDP.pdfGrok 4.6 xhigh: 17.2 on GDP.pdfGLM 5.3 max: 11.2 on GDP.pdfDeepSeek V4.1 Flash 1017 max: 17.2 on GDP.pdfHy4: 21.6 on GDP.pdfLing 3.1 Flash: 12.6 on GDP.pdfGemini 3.8 Flash high: 21.0 on GDP.pdfStep 5 Preview: 14.8 on GDP.pdfDeepSeek V4.1 Flash max: 12.8 on GDP.pdfQwen 3.8 2.4T-A95B: 15.6 on GDP.pdfMiMo V2.6 Pro: 19.2 on GDP.pdfQwen 3.8 Flash: 15.6 on GDP.pdfKimi K3 max: 22.0 on GDP.pdfGLM 5.3 Flash max: 15.4 on GDP.pdfGPT-5.6 Terra max: 24.0 on GDP.pdfMistral Large 4 high: 17.6 on GDP.pdfMuse Spark Open xhigh: 18.6 on GDP.pdfJoyAI-VL-Preview max: 20.6 on GDP.pdfMiMo V2.6 Flash: 9.0 on GDP.pdfSonnet 5 max: 13.2 on GDP.pdfGPT-6 Luna max: 22.8 on GDP.pdfMiniMax M3.1 max: 20.8 on GDP.pdfAgnes 2.5 Pro Theta: 14.2 on GDP.pdfGLM 5.2 max: 10.4 on GDP.pdfNanbeige4.3-33B: 11.2 on GDP.pdfQwen 3.8 27B xhigh: 16.4 on GDP.pdfInkling 2: 16.8 on GDP.pdfGPT-5.6 Luna max: 24.0 on GDP.pdfK2 Horizon 375B A23B high: 7.4 on GDP.pdfDeepSeek V4 Pro 0813 max: 11.4 on GDP.pdfBeam 1.0: 11.6 on GDP.pdfJT-4.1 Flash 236B A21B: 6.6 on GDP.pdfSolar Pro 4: 7.0 on GDP.pdfMiniMax M3: 9.8 on GDP.pdfK2 Horizon MoVA 36B A4B: 4.2 on GDP.pdfNemotron 3.5 Super VL high: 8.0 on GDP.pdfK2 Horizon 32B: 3.8 on GDP.pdfNemotron 3 Ultra max: 5.0 on GDP.pdfSolar Mini 4: 5.0 on GDP.pdfGemini 3.5 Flash Lite high: 13.6 on GDP.pdfInkling Small: 11.2 on GDP.pdfApodex 1.1: 11.2 on GDP.pdfK2 Horizon 7B: 3.2 on GDP.pdfInkling xhigh: 12.8 on GDP.pdfA.X-K2: 6.4 on GDP.pdfK-EXAONE 2.0: 4.2 on GDP.pdfLFM3-1.7B: 0.4 on GDP.pdfK2 Horizon 3.7B: 3.2 on GDP.pdfGranite 4.2 30B: 1.8 on GDP.pdfNemotron 3.5 Lightning: 3.0 on GDP.pdfGranite 4.2 8B: 0.6 on GDP.pdfGranite 4.2 3B: 0.2 on GDP.pdfMuse Glimmer high: 10.0 on GDP.pdfLing 3.0 Tiny: 0.8 on GDP.pdfNorth Mini Code: 1.8 on GDP.pdfK2 Horizon 0.9B: 0.8 on GDP.pdfLFM3-VL-2B: 0.0 on GDP.pdf
0255075100-2-101Score on the benchmarkAbility (0 is the middle of the models Priors tracks)Top model todayGemini 4 Neon xhigh: 79.3 on AutomationBench-AAFable 5.2 max: 76.4 on AutomationBench-AAGemini 4 Carbon high: 77.8 on AutomationBench-AAGPT-6.2 Astra max: 77.2 on AutomationBench-AAOpus 5.5 max: 69.5 on AutomationBench-AAGemini 4 Hydrogen xhigh: 76.1 on AutomationBench-AAGPT-6 Astra max: 68.5 on AutomationBench-AAGPT-6.2 Sol max: 71.3 on AutomationBench-AASonnet 5.5 max: 71.8 on AutomationBench-AAGPT-6.1 Sol max: 64.9 on AutomationBench-AAGemini 4 Argon high: 77.5 on AutomationBench-AAMuse Wonder Full max: 71.3 on AutomationBench-AAGPT-6.1 Terra max: 60.2 on AutomationBench-AAFable 5.1 xhigh: 57.8 on AutomationBench-AAKimi K3.1 max: 73.3 on AutomationBench-AAMuse Wonder max: 68.1 on AutomationBench-AAQwen 4 Max xhigh: 72.6 on AutomationBench-AAOpus 5 max: 56.6 on AutomationBench-AAGPT-6 Sol max: 61.6 on AutomationBench-AAGPT-5.6 Sol max: 60.1 on AutomationBench-AAMuse Ember max: 68.1 on AutomationBench-AAMuse Spark 1.3 max: 57.9 on AutomationBench-AAGrok 4.7 high: 63.5 on AutomationBench-AAHaiku 5.5 max: 35.4 on AutomationBench-AAQwen 3.8 Max xhigh: 56.2 on AutomationBench-AAInkling 2 Small: 45.6 on AutomationBench-AAGrok 4.6 xhigh: 67.0 on AutomationBench-AAGLM 5.3 max: 62.2 on AutomationBench-AADeepSeek V4.1 Flash 1017 max: 66.6 on AutomationBench-AAHy4: 65.6 on AutomationBench-AALing 3.1 Flash: 67.5 on AutomationBench-AAGemini 3.8 Flash high: 59.9 on AutomationBench-AAStep 5 Preview: 51.0 on AutomationBench-AADeepSeek V4.1 Flash max: 68.9 on AutomationBench-AAQwen 3.8 2.4T-A95B: 57.2 on AutomationBench-AAMiMo V2.6 Pro: 58.6 on AutomationBench-AAQwen 3.8 Flash: 55.9 on AutomationBench-AAKimi K3 max: 58.3 on AutomationBench-AAGLM 5.3 Flash max: 60.4 on AutomationBench-AAGPT-5.6 Terra max: 59.6 on AutomationBench-AAMistral Large 4 high: 57.8 on AutomationBench-AAMuse Spark Open xhigh: 49.7 on AutomationBench-AAJoyAI-VL-Preview max: 61.7 on AutomationBench-AAMiMo V2.6 Flash: 64.1 on AutomationBench-AASonnet 5 max: 36.5 on AutomationBench-AAGPT-6 Luna max: 53.2 on AutomationBench-AAMiniMax M3.1 max: 72.5 on AutomationBench-AAAgnes 2.5 Pro Theta: 48.8 on AutomationBench-AAGLM 5.2 max: 28.4 on AutomationBench-AANanbeige4.3-33B: 53.1 on AutomationBench-AAQwen 3.8 27B xhigh: 48.2 on AutomationBench-AAInkling 2: 41.8 on AutomationBench-AAGPT-5.6 Luna max: 50.2 on AutomationBench-AAK2 Horizon 375B A23B high: 37.2 on AutomationBench-AADeepSeek V4 Pro 0813 max: 56.7 on AutomationBench-AABeam 1.0: 48.5 on AutomationBench-AAJT-4.1 Flash 236B A21B: 36.7 on AutomationBench-AASolar Pro 4: 13.6 on AutomationBench-AAMiniMax M3: 21.3 on AutomationBench-AAK2 Horizon MoVA 36B A4B: 26.1 on AutomationBench-AANemotron 3.5 Super VL high: 14.8 on AutomationBench-AAK2 Horizon 32B: 29.4 on AutomationBench-AANemotron 3 Ultra max: 3.0 on AutomationBench-AASolar Mini 4: 22.3 on AutomationBench-AAGemini 3.5 Flash Lite high: 25.0 on AutomationBench-AAInkling Small: 4.8 on AutomationBench-AAApodex 1.1: 28.2 on AutomationBench-AAK2 Horizon 7B: 20.9 on AutomationBench-AAInkling xhigh: 5.0 on AutomationBench-AAA.X-K2: 2.5 on AutomationBench-AAK-EXAONE 2.0: 3.9 on AutomationBench-AALFM3-1.7B: 7.0 on AutomationBench-AAK2 Horizon 3.7B: 14.7 on AutomationBench-AAGranite 4.2 30B: 1.5 on AutomationBench-AANemotron 3.5 Lightning: 0.8 on AutomationBench-AAGranite 4.2 8B: 2.1 on AutomationBench-AAGranite 4.2 3B: 0.4 on AutomationBench-AAMuse Glimmer high: 6.8 on AutomationBench-AALing 3.0 Tiny: 0.3 on AutomationBench-AANorth Mini Code: 1.5 on AutomationBench-AAK2 Horizon 0.9B: 0.2 on AutomationBench-AALFM3-VL-2B: 0.2 on AutomationBench-AAGemini 4 Neon xhigh: 70.0 on Terminal-Bench-Science-AAFable 5.2 max: 73.3 on Terminal-Bench-Science-AAGemini 4 Carbon high: 64.8 on Terminal-Bench-Science-AAGPT-6.2 Astra max: 78.1 on Terminal-Bench-Science-AAOpus 5.5 max: 59.0 on Terminal-Bench-Science-AAGemini 4 Hydrogen xhigh: 49.5 on Terminal-Bench-Science-AAGPT-6 Astra max: 63.3 on Terminal-Bench-Science-AAGPT-6.2 Sol max: 67.6 on Terminal-Bench-Science-AASonnet 5.5 max: 53.3 on Terminal-Bench-Science-AAGPT-6.1 Sol max: 58.1 on Terminal-Bench-Science-AAGemini 4 Argon high: 52.9 on Terminal-Bench-Science-AAMuse Wonder Full max: 37.1 on Terminal-Bench-Science-AAGPT-6.1 Terra max: 57.6 on Terminal-Bench-Science-AAFable 5.1 xhigh: 40.5 on Terminal-Bench-Science-AAKimi K3.1 max: 34.3 on Terminal-Bench-Science-AAMuse Wonder max: 25.7 on Terminal-Bench-Science-AAQwen 4 Max xhigh: 23.3 on Terminal-Bench-Science-AAOpus 5 max: 28.6 on Terminal-Bench-Science-AAGPT-6 Sol max: 30.0 on Terminal-Bench-Science-AAGPT-5.6 Sol max: 22.4 on Terminal-Bench-Science-AAMuse Ember max: 20.0 on Terminal-Bench-Science-AAMuse Spark 1.3 max: 11.0 on Terminal-Bench-Science-AAGrok 4.7 high: 9.5 on Terminal-Bench-Science-AAHaiku 5.5 max: 20.0 on Terminal-Bench-Science-AAQwen 3.8 Max xhigh: 11.9 on Terminal-Bench-Science-AAInkling 2 Small: 5.7 on Terminal-Bench-Science-AAGrok 4.6 xhigh: 2.9 on Terminal-Bench-Science-AAGLM 5.3 max: 9.5 on Terminal-Bench-Science-AADeepSeek V4.1 Flash 1017 max: 12.9 on Terminal-Bench-Science-AAHy4: 12.9 on Terminal-Bench-Science-AALing 3.1 Flash: 4.3 on Terminal-Bench-Science-AAGemini 3.8 Flash high: 10.0 on Terminal-Bench-Science-AAStep 5 Preview: 2.4 on Terminal-Bench-Science-AADeepSeek V4.1 Flash max: 9.0 on Terminal-Bench-Science-AAMiMo V2.6 Pro: 5.7 on Terminal-Bench-Science-AAQwen 3.8 Flash: 7.1 on Terminal-Bench-Science-AAKimi K3 max: 8.1 on Terminal-Bench-Science-AAGLM 5.3 Flash max: 4.8 on Terminal-Bench-Science-AAGPT-5.6 Terra max: 13.8 on Terminal-Bench-Science-AAMistral Large 4 high: 4.3 on Terminal-Bench-Science-AAMuse Spark Open xhigh: 1.9 on Terminal-Bench-Science-AAJoyAI-VL-Preview max: 3.8 on Terminal-Bench-Science-AAMiMo V2.6 Flash: 4.3 on Terminal-Bench-Science-AAGPT-6 Luna max: 8.6 on Terminal-Bench-Science-AAMiniMax M3.1 max: 3.3 on Terminal-Bench-Science-AAAgnes 2.5 Pro Theta: 3.8 on Terminal-Bench-Science-AANanbeige4.3-33B: 1.0 on Terminal-Bench-Science-AAQwen 3.8 27B xhigh: 0.5 on Terminal-Bench-Science-AAInkling 2: 1.4 on Terminal-Bench-Science-AAGPT-5.6 Luna max: 3.3 on Terminal-Bench-Science-AAK2 Horizon 375B A23B high: 0.0 on Terminal-Bench-Science-AADeepSeek V4 Pro 0813 max: 5.7 on Terminal-Bench-Science-AABeam 1.0: 0.5 on Terminal-Bench-Science-AAJT-4.1 Flash 236B A21B: 0.0 on Terminal-Bench-Science-AASolar Pro 4: 0.0 on Terminal-Bench-Science-AAMiniMax M3: 0.5 on Terminal-Bench-Science-AAK2 Horizon MoVA 36B A4B: 0.0 on Terminal-Bench-Science-AANemotron 3.5 Super VL high: 0.0 on Terminal-Bench-Science-AAK2 Horizon 32B: 0.0 on Terminal-Bench-Science-AANemotron 3 Ultra max: 0.0 on Terminal-Bench-Science-AAGemini 3.5 Flash Lite high: 0.0 on Terminal-Bench-Science-AAInkling Small: 0.0 on Terminal-Bench-Science-AAApodex 1.1: 0.5 on Terminal-Bench-Science-AAK2 Horizon 7B: 0.0 on Terminal-Bench-Science-AAInkling xhigh: 0.0 on Terminal-Bench-Science-AAA.X-K2: 0.0 on Terminal-Bench-Science-AAK-EXAONE 2.0: 0.0 on Terminal-Bench-Science-AALFM3-1.7B: 0.0 on Terminal-Bench-Science-AAK2 Horizon 3.7B: 0.0 on Terminal-Bench-Science-AAGranite 4.2 30B: 0.0 on Terminal-Bench-Science-AANemotron 3.5 Lightning: 0.0 on Terminal-Bench-Science-AAGranite 4.2 8B: 0.0 on Terminal-Bench-Science-AAGranite 4.2 3B: 0.0 on Terminal-Bench-Science-AAMuse Glimmer high: 0.0 on Terminal-Bench-Science-AANorth Mini Code: 0.0 on Terminal-Bench-Science-AAK2 Horizon 0.9B: 0.0 on Terminal-Bench-Science-AALFM3-VL-2B: 0.0 on Terminal-Bench-Science-AAGemini 4 Neon xhigh: 28.0 on GDP.pdfFable 5.2 max: 24.2 on GDP.pdfGemini 4 Carbon high: 28.4 on GDP.pdfGPT-6.2 Astra max: 29.8 on GDP.pdfOpus 5.5 max: 26.2 on GDP.pdfGemini 4 Hydrogen xhigh: 27.6 on GDP.pdfGPT-6 Astra max: 31.0 on GDP.pdfGPT-6.2 Sol max: 30.0 on GDP.pdfSonnet 5.5 max: 25.8 on GDP.pdfGPT-6.1 Sol max: 31.0 on GDP.pdfGemini 4 Argon high: 21.8 on GDP.pdfMuse Wonder Full max: 32.6 on GDP.pdfGPT-6.1 Terra max: 28.6 on GDP.pdfFable 5.1 xhigh: 26.2 on GDP.pdfKimi K3.1 max: 31.4 on GDP.pdfMuse Wonder max: 30.6 on GDP.pdfQwen 4 Max xhigh: 26.2 on GDP.pdfOpus 5 max: 21.6 on GDP.pdfGPT-6 Sol max: 25.2 on GDP.pdfGPT-5.6 Sol max: 27.2 on GDP.pdfMuse Ember max: 22.8 on GDP.pdfMuse Spark 1.3 max: 26.6 on GDP.pdfGrok 4.7 high: 23.2 on GDP.pdfHaiku 5.5 max: 20.8 on GDP.pdfQwen 3.8 Max xhigh: 22.8 on GDP.pdfInkling 2 Small: 25.2 on GDP.pdfGrok 4.6 xhigh: 17.2 on GDP.pdfGLM 5.3 max: 11.2 on GDP.pdfDeepSeek V4.1 Flash 1017 max: 17.2 on GDP.pdfHy4: 21.6 on GDP.pdfLing 3.1 Flash: 12.6 on GDP.pdfGemini 3.8 Flash high: 21.0 on GDP.pdfStep 5 Preview: 14.8 on GDP.pdfDeepSeek V4.1 Flash max: 12.8 on GDP.pdfQwen 3.8 2.4T-A95B: 15.6 on GDP.pdfMiMo V2.6 Pro: 19.2 on GDP.pdfQwen 3.8 Flash: 15.6 on GDP.pdfKimi K3 max: 22.0 on GDP.pdfGLM 5.3 Flash max: 15.4 on GDP.pdfGPT-5.6 Terra max: 24.0 on GDP.pdfMistral Large 4 high: 17.6 on GDP.pdfMuse Spark Open xhigh: 18.6 on GDP.pdfJoyAI-VL-Preview max: 20.6 on GDP.pdfMiMo V2.6 Flash: 9.0 on GDP.pdfSonnet 5 max: 13.2 on GDP.pdfGPT-6 Luna max: 22.8 on GDP.pdfMiniMax M3.1 max: 20.8 on GDP.pdfAgnes 2.5 Pro Theta: 14.2 on GDP.pdfGLM 5.2 max: 10.4 on GDP.pdfNanbeige4.3-33B: 11.2 on GDP.pdfQwen 3.8 27B xhigh: 16.4 on GDP.pdfInkling 2: 16.8 on GDP.pdfGPT-5.6 Luna max: 24.0 on GDP.pdfK2 Horizon 375B A23B high: 7.4 on GDP.pdfDeepSeek V4 Pro 0813 max: 11.4 on GDP.pdfBeam 1.0: 11.6 on GDP.pdfJT-4.1 Flash 236B A21B: 6.6 on GDP.pdfSolar Pro 4: 7.0 on GDP.pdfMiniMax M3: 9.8 on GDP.pdfK2 Horizon MoVA 36B A4B: 4.2 on GDP.pdfNemotron 3.5 Super VL high: 8.0 on GDP.pdfK2 Horizon 32B: 3.8 on GDP.pdfNemotron 3 Ultra max: 5.0 on GDP.pdfSolar Mini 4: 5.0 on GDP.pdfGemini 3.5 Flash Lite high: 13.6 on GDP.pdfInkling Small: 11.2 on GDP.pdfApodex 1.1: 11.2 on GDP.pdfK2 Horizon 7B: 3.2 on GDP.pdfInkling xhigh: 12.8 on GDP.pdfA.X-K2: 6.4 on GDP.pdfK-EXAONE 2.0: 4.2 on GDP.pdfLFM3-1.7B: 0.4 on GDP.pdfK2 Horizon 3.7B: 3.2 on GDP.pdfGranite 4.2 30B: 1.8 on GDP.pdfNemotron 3.5 Lightning: 3.0 on GDP.pdfGranite 4.2 8B: 0.6 on GDP.pdfGranite 4.2 3B: 0.2 on GDP.pdfMuse Glimmer high: 10.0 on GDP.pdfLing 3.0 Tiny: 0.8 on GDP.pdfNorth Mini Code: 1.8 on GDP.pdfK2 Horizon 0.9B: 0.8 on GDP.pdfLFM3-VL-2B: 0.0 on GDP.pdf

How an ability becomes a 0 to 100 score

An ability is a position along the bottom of that chart: a number like 0.75, with no top or bottom of its own. Priors does not stretch or rescale it onto 0 to 100. It turns it back into benchmark scores, in two steps.

  1. Read every benchmark’s curve at the model’s ability. That gives the score the curve expects from a model that able.
  2. Average those expected scores with your weights. This is the same sum the weighted average does, on expected scores instead of real ones.
Gemini 4 Neon xhigh, step by stepIts fitted ability is 0.75. Each small chart is one benchmark's curve from the chart above. The dot is where the curve is read, and the ring is the score the model really got.
Its curve, read at 0.75Curve saysReally gotWeightAdds
Terminal-Bench-Science-AA73.1 × 15%70.0× 15%10.97
AA-Briefcase v1.163.8 × 15%59.9× 15%9.57
CritPt v2 Hard Subset74.6 × 10%60.5× 10%7.46
FrontierCode 2.0-AA48.0 × 15%44.4× 15%7.20
Terminal-Bench v4-AA71.1 × 10%63.6× 10%7.11
HLE-Diamond66.2 × 10%71.3× 10%6.62
AA-Omniscience v1.1 Accuracy64.5 × 10%60.5× 10%6.45
AutomationBench-AA80.9 × 5%79.3× 5%4.04
AA-Omniscience v1.1 Non-hallucination59.2 × 5%85.6× 5%2.96
GDP.pdf35.5 × 5%28.0× 5%1.78
Added up: the IRT ability score64.2
The same sum on the scores it really got: the weighted average61.4

That is why the score cannot leave 0 to 100. Every reading is a benchmark score, so it lies between 0 and 100, and an average of them does too. A score of 100 would mean the curves expect full marks on every benchmark in the mix. It is also why the score reads like the default one: it is the same weighted average, with each real score replaced by what the model’s ability predicts. The likely range is converted the same way, by reading the curves at both ends of the ability’s range.

Your weights also decide how the ability itself is found. A benchmark’s weight is its share of the pull on ability, so a steep benchmark does not pull harder than its weight says.

When it differs. A model that does well on hard benchmarks and loses a few points on easy ones moves up. A model whose average is carried by one unusual result moves down, because one ability cannot explain an outlier. Nothing is filled in: a benchmark a model never ran is left to its curve.

Limits. One ability per model is a simplification, so a model that is strong at one kind of task and weak at another is described less well. The ability starts from how the model does across every benchmark Priors has, not only your mix, so on a mix of one or two benchmarks it follows them less closely than the average does. Each score has a likely range, and the ranges of neighbouring models usually overlap.

ECI-style

Epoch AI publishes the Epoch Capabilities Index: one capability number per model, fitted across many benchmarks at once. Each benchmark gets a difficulty and a slope, each model a capability, and a model’s expected score on a benchmark is sigmoid(slope × (capability − difficulty)). Capabilities are then rescaled in a straight line onto a scale with no upper limit.

The same curves on an open scaleThe same mix under the ECI-style fit, on 2026-11-22. Capability and difficulty share one axis: GDP.pdf crosses halfway at 185, past every model today. A ring is Epoch's own ECI for the same model, where Epoch has scored it.
AutomationBench-AA halfway at 157Terminal-Bench-Science-AA halfway at 167GDP.pdf halfway at 185the mix’s other 7
0255075100110120130140150160170180190Score on the benchmarkECI-style scoreTop model todayFable 5.2 max: 76.4 on AutomationBench-AAGPT-6.2 Astra max: 77.2 on AutomationBench-AAGemini 4 Neon xhigh: 79.3 on AutomationBench-AAGemini 4 Carbon high: 77.8 on AutomationBench-AAGPT-6.2 Sol max: 71.3 on AutomationBench-AAOpus 5.5 max: 69.5 on AutomationBench-AAGPT-6 Astra max: 68.5 on AutomationBench-AAGPT-6.1 Sol max: 64.9 on AutomationBench-AASonnet 5.5 max: 71.8 on AutomationBench-AAGPT-6.1 Terra max: 60.2 on AutomationBench-AAGemini 4 Hydrogen xhigh: 76.1 on AutomationBench-AAGemini 4 Argon high: 77.5 on AutomationBench-AAFable 5.1 xhigh: 57.8 on AutomationBench-AAMuse Wonder Full max: 71.3 on AutomationBench-AAKimi K3.1 max: 73.3 on AutomationBench-AAGPT-6 Sol max: 61.6 on AutomationBench-AAOpus 5 max: 56.6 on AutomationBench-AAMuse Wonder max: 68.1 on AutomationBench-AAQwen 4 Max xhigh: 72.6 on AutomationBench-AAGPT-5.6 Sol max: 60.1 on AutomationBench-AAHy4: 65.6 on AutomationBench-AAHaiku 5.5 max: 35.4 on AutomationBench-AAMuse Ember max: 68.1 on AutomationBench-AAMuse Spark 1.3 max: 57.9 on AutomationBench-AAGPT-5.6 Terra max: 59.6 on AutomationBench-AAGrok 4.7 xhigh: 65.6 on AutomationBench-AADeepSeek V4.1 Flash 1017 max: 66.6 on AutomationBench-AAQwen 3.8 Max xhigh: 56.2 on AutomationBench-AAGemini 3.8 Flash high: 59.9 on AutomationBench-AAGLM 5.3 max: 62.2 on AutomationBench-AADeepSeek V4.1 Flash max: 68.9 on AutomationBench-AAMiniMax M3.1 max: 72.5 on AutomationBench-AAGrok 4.6 high: 66.7 on AutomationBench-AAKimi K3 max: 58.3 on AutomationBench-AAMiMo V2.6 Pro: 58.6 on AutomationBench-AALing 3.1 Flash: 67.5 on AutomationBench-AAInkling 2 Small: 45.6 on AutomationBench-AAGPT-6 Luna max: 53.2 on AutomationBench-AAJoyAI-VL-Preview max: 61.7 on AutomationBench-AAStep 5 Preview: 51.0 on AutomationBench-AAQwen 3.8 Flash: 55.9 on AutomationBench-AAGLM 5.3 Flash max: 60.4 on AutomationBench-AAMiMo V2.6 Flash: 64.1 on AutomationBench-AAQwen 3.8 2.4T-A95B: 57.2 on AutomationBench-AAMistral Large 4 high: 57.8 on AutomationBench-AAGPT-5.6 Luna max: 50.2 on AutomationBench-AADeepSeek V4 Pro 0813 max: 56.7 on AutomationBench-AASonnet 5 max: 36.5 on AutomationBench-AAMuse Spark Open xhigh: 49.7 on AutomationBench-AAAgnes 2.5 Pro Theta: 48.8 on AutomationBench-AAMiniMax M3: 21.3 on AutomationBench-AAJT-4.1 Flash 236B A21B: 36.7 on AutomationBench-AAInkling 2: 41.8 on AutomationBench-AANanbeige4.3-33B: 53.1 on AutomationBench-AAQwen 3.8 27B xhigh: 48.2 on AutomationBench-AABeam 1.0: 48.5 on AutomationBench-AAGLM 5.2 max: 28.4 on AutomationBench-AAK2 Horizon 375B A23B high: 37.2 on AutomationBench-AAApodex 1.1: 28.2 on AutomationBench-AAK2 Horizon 32B: 29.4 on AutomationBench-AAK2 Horizon MoVA 36B A4B: 26.1 on AutomationBench-AASolar Pro 4: 13.6 on AutomationBench-AAInkling Small: 4.8 on AutomationBench-AAInkling xhigh: 5.0 on AutomationBench-AANemotron 3.5 Super VL high: 14.8 on AutomationBench-AASolar Mini 4: 22.3 on AutomationBench-AANemotron 3 Ultra max: 3.0 on AutomationBench-AAGemini 3.5 Flash Lite high: 25.0 on AutomationBench-AAK2 Horizon 7B: 20.9 on AutomationBench-AAA.X-K2: 2.5 on AutomationBench-AAK2 Horizon 3.7B: 14.7 on AutomationBench-AAK-EXAONE 2.0: 3.9 on AutomationBench-AALing 3.0 Tiny: 0.3 on AutomationBench-AAMuse Glimmer high: 6.8 on AutomationBench-AALFM3-1.7B: 7.0 on AutomationBench-AANemotron 3.5 Lightning: 0.8 on AutomationBench-AAGranite 4.2 30B: 1.5 on AutomationBench-AAGranite 4.2 8B: 2.1 on AutomationBench-AAGranite 4.2 3B: 0.4 on AutomationBench-AAFable 5.2 max: 73.3 on Terminal-Bench-Science-AAGPT-6.2 Astra max: 78.1 on Terminal-Bench-Science-AAGemini 4 Neon xhigh: 70.0 on Terminal-Bench-Science-AAGemini 4 Carbon high: 64.8 on Terminal-Bench-Science-AAGPT-6.2 Sol max: 67.6 on Terminal-Bench-Science-AAOpus 5.5 max: 59.0 on Terminal-Bench-Science-AAGPT-6 Astra max: 63.3 on Terminal-Bench-Science-AAGPT-6.1 Sol max: 58.1 on Terminal-Bench-Science-AASonnet 5.5 max: 53.3 on Terminal-Bench-Science-AAGPT-6.1 Terra max: 57.6 on Terminal-Bench-Science-AAGemini 4 Hydrogen xhigh: 49.5 on Terminal-Bench-Science-AAGemini 4 Argon high: 52.9 on Terminal-Bench-Science-AAFable 5.1 xhigh: 40.5 on Terminal-Bench-Science-AAMuse Wonder Full max: 37.1 on Terminal-Bench-Science-AAKimi K3.1 max: 34.3 on Terminal-Bench-Science-AAGPT-6 Sol max: 30.0 on Terminal-Bench-Science-AAOpus 5 max: 28.6 on Terminal-Bench-Science-AAMuse Wonder max: 25.7 on Terminal-Bench-Science-AAQwen 4 Max xhigh: 23.3 on Terminal-Bench-Science-AAGPT-5.6 Sol max: 22.4 on Terminal-Bench-Science-AAHy4: 12.9 on Terminal-Bench-Science-AAHaiku 5.5 max: 20.0 on Terminal-Bench-Science-AAMuse Ember max: 20.0 on Terminal-Bench-Science-AAMuse Spark 1.3 max: 11.0 on Terminal-Bench-Science-AAGPT-5.6 Terra max: 13.8 on Terminal-Bench-Science-AAGrok 4.7 xhigh: 11.4 on Terminal-Bench-Science-AADeepSeek V4.1 Flash 1017 max: 12.9 on Terminal-Bench-Science-AAQwen 3.8 Max xhigh: 11.9 on Terminal-Bench-Science-AAGemini 3.8 Flash high: 10.0 on Terminal-Bench-Science-AAGLM 5.3 max: 9.5 on Terminal-Bench-Science-AADeepSeek V4.1 Flash max: 9.0 on Terminal-Bench-Science-AAMiniMax M3.1 max: 3.3 on Terminal-Bench-Science-AAGrok 4.6 high: 6.2 on Terminal-Bench-Science-AAKimi K3 max: 8.1 on Terminal-Bench-Science-AAMiMo V2.6 Pro: 5.7 on Terminal-Bench-Science-AALing 3.1 Flash: 4.3 on Terminal-Bench-Science-AAInkling 2 Small: 5.7 on Terminal-Bench-Science-AAGPT-6 Luna max: 8.6 on Terminal-Bench-Science-AAJoyAI-VL-Preview max: 3.8 on Terminal-Bench-Science-AAStep 5 Preview: 2.4 on Terminal-Bench-Science-AAQwen 3.8 Flash: 7.1 on Terminal-Bench-Science-AAGLM 5.3 Flash max: 4.8 on Terminal-Bench-Science-AAMiMo V2.6 Flash: 4.3 on Terminal-Bench-Science-AAMistral Large 4 high: 4.3 on Terminal-Bench-Science-AAGPT-5.6 Luna max: 3.3 on Terminal-Bench-Science-AADeepSeek V4 Pro 0813 max: 5.7 on Terminal-Bench-Science-AAMuse Spark Open xhigh: 1.9 on Terminal-Bench-Science-AAAgnes 2.5 Pro Theta: 3.8 on Terminal-Bench-Science-AAMiniMax M3: 0.5 on Terminal-Bench-Science-AAJT-4.1 Flash 236B A21B: 0.0 on Terminal-Bench-Science-AAInkling 2: 1.4 on Terminal-Bench-Science-AANanbeige4.3-33B: 1.0 on Terminal-Bench-Science-AAQwen 3.8 27B xhigh: 0.5 on Terminal-Bench-Science-AABeam 1.0: 0.5 on Terminal-Bench-Science-AAK2 Horizon 375B A23B high: 0.0 on Terminal-Bench-Science-AAApodex 1.1: 0.5 on Terminal-Bench-Science-AAK2 Horizon 32B: 0.0 on Terminal-Bench-Science-AAK2 Horizon MoVA 36B A4B: 0.0 on Terminal-Bench-Science-AASolar Pro 4: 0.0 on Terminal-Bench-Science-AAInkling Small: 0.0 on Terminal-Bench-Science-AAInkling xhigh: 0.0 on Terminal-Bench-Science-AANemotron 3.5 Super VL high: 0.0 on Terminal-Bench-Science-AANemotron 3 Ultra max: 0.0 on Terminal-Bench-Science-AAGemini 3.5 Flash Lite high: 0.0 on Terminal-Bench-Science-AAK2 Horizon 7B: 0.0 on Terminal-Bench-Science-AAA.X-K2: 0.0 on Terminal-Bench-Science-AAK2 Horizon 3.7B: 0.0 on Terminal-Bench-Science-AAK-EXAONE 2.0: 0.0 on Terminal-Bench-Science-AAMuse Glimmer high: 0.0 on Terminal-Bench-Science-AALFM3-1.7B: 0.0 on Terminal-Bench-Science-AANemotron 3.5 Lightning: 0.0 on Terminal-Bench-Science-AAGranite 4.2 30B: 0.0 on Terminal-Bench-Science-AAGranite 4.2 8B: 0.0 on Terminal-Bench-Science-AAGranite 4.2 3B: 0.0 on Terminal-Bench-Science-AAFable 5.2 max: 24.2 on GDP.pdfGPT-6.2 Astra max: 29.8 on GDP.pdfGemini 4 Neon xhigh: 28.0 on GDP.pdfGemini 4 Carbon high: 28.4 on GDP.pdfGPT-6.2 Sol max: 30.0 on GDP.pdfOpus 5.5 max: 26.2 on GDP.pdfGPT-6 Astra max: 31.0 on GDP.pdfGPT-6.1 Sol max: 31.0 on GDP.pdfSonnet 5.5 max: 25.8 on GDP.pdfGPT-6.1 Terra max: 28.6 on GDP.pdfGemini 4 Hydrogen xhigh: 27.6 on GDP.pdfGemini 4 Argon high: 21.8 on GDP.pdfFable 5.1 xhigh: 26.2 on GDP.pdfMuse Wonder Full max: 32.6 on GDP.pdfKimi K3.1 max: 31.4 on GDP.pdfGPT-6 Sol max: 25.2 on GDP.pdfOpus 5 max: 21.6 on GDP.pdfMuse Wonder max: 30.6 on GDP.pdfQwen 4 Max xhigh: 26.2 on GDP.pdfGPT-5.6 Sol max: 27.2 on GDP.pdfHy4: 21.6 on GDP.pdfHaiku 5.5 max: 20.8 on GDP.pdfMuse Ember max: 22.8 on GDP.pdfMuse Spark 1.3 max: 26.6 on GDP.pdfGPT-5.6 Terra max: 24.0 on GDP.pdfGrok 4.7 xhigh: 20.0 on GDP.pdfDeepSeek V4.1 Flash 1017 max: 17.2 on GDP.pdfQwen 3.8 Max xhigh: 22.8 on GDP.pdfGemini 3.8 Flash high: 21.0 on GDP.pdfGLM 5.3 max: 11.2 on GDP.pdfDeepSeek V4.1 Flash max: 12.8 on GDP.pdfMiniMax M3.1 max: 20.8 on GDP.pdfGrok 4.6 high: 17.0 on GDP.pdfKimi K3 max: 22.0 on GDP.pdfMiMo V2.6 Pro: 19.2 on GDP.pdfLing 3.1 Flash: 12.6 on GDP.pdfInkling 2 Small: 25.2 on GDP.pdfGPT-6 Luna max: 22.8 on GDP.pdfJoyAI-VL-Preview max: 20.6 on GDP.pdfStep 5 Preview: 14.8 on GDP.pdfQwen 3.8 Flash: 15.6 on GDP.pdfGLM 5.3 Flash max: 15.4 on GDP.pdfMiMo V2.6 Flash: 9.0 on GDP.pdfQwen 3.8 2.4T-A95B: 15.6 on GDP.pdfMistral Large 4 high: 17.6 on GDP.pdfGPT-5.6 Luna max: 24.0 on GDP.pdfDeepSeek V4 Pro 0813 max: 11.4 on GDP.pdfSonnet 5 max: 13.2 on GDP.pdfMuse Spark Open xhigh: 18.6 on GDP.pdfAgnes 2.5 Pro Theta: 14.2 on GDP.pdfMiniMax M3: 9.8 on GDP.pdfJT-4.1 Flash 236B A21B: 6.6 on GDP.pdfInkling 2: 16.8 on GDP.pdfNanbeige4.3-33B: 11.2 on GDP.pdfQwen 3.8 27B xhigh: 16.4 on GDP.pdfBeam 1.0: 11.6 on GDP.pdfGLM 5.2 max: 10.4 on GDP.pdfK2 Horizon 375B A23B high: 7.4 on GDP.pdfApodex 1.1: 11.2 on GDP.pdfK2 Horizon 32B: 3.8 on GDP.pdfK2 Horizon MoVA 36B A4B: 4.2 on GDP.pdfSolar Pro 4: 7.0 on GDP.pdfInkling Small: 11.2 on GDP.pdfInkling xhigh: 12.8 on GDP.pdfNemotron 3.5 Super VL high: 8.0 on GDP.pdfSolar Mini 4: 5.0 on GDP.pdfNemotron 3 Ultra max: 5.0 on GDP.pdfGemini 3.5 Flash Lite high: 13.6 on GDP.pdfK2 Horizon 7B: 3.2 on GDP.pdfA.X-K2: 6.4 on GDP.pdfK2 Horizon 3.7B: 3.2 on GDP.pdfK-EXAONE 2.0: 4.2 on GDP.pdfLing 3.0 Tiny: 0.8 on GDP.pdfMuse Glimmer high: 10.0 on GDP.pdfLFM3-1.7B: 0.4 on GDP.pdfNemotron 3.5 Lightning: 3.0 on GDP.pdfGranite 4.2 30B: 1.8 on GDP.pdfGranite 4.2 8B: 0.6 on GDP.pdfGranite 4.2 3B: 0.2 on GDP.pdfThe top models on the same axis, each with its likely rangeEpoch's own ECI for Fable 5.2 max: 173.4Fable 5.2 max: 171.8, likely 169.1 to 174.5Fable 5.2 max 171.8Epoch's own ECI for GPT-6.2 Astra max: 172.8GPT-6.2 Astra max: 171.5, likely 168.9 to 174.1GPT-6.2 Astra max 171.5Gemini 4 Neon xhigh: 170.3, likely 167.9 to 172.8Gemini 4 Neon xhigh 170.3Gemini 4 Carbon high: 169.7, likely 167.4 to 172.1Gemini 4 Carbon high 169.7GPT-6.2 Sol max: 169.6, likely 167.2 to 171.9GPT-6.2 Sol max 169.6Epoch's own ECI for Opus 5.5 max: 168.7Opus 5.5 max: 169.3, likely 167.0 to 171.6Opus 5.5 max 169.3Epoch's own ECI for GPT-6 Astra max: 169.2GPT-6 Astra max: 169.2, likely 166.9 to 171.5GPT-6 Astra max 169.2Epoch's own ECI for GPT-6.1 Sol max: 168.6GPT-6.1 Sol max: 168.3, likely 166.1 to 170.5GPT-6.1 Sol max 168.3Epoch's own ECI for Sonnet 5.5 max: 166.6Sonnet 5.5 max: 167.9, likely 165.7 to 170.1Sonnet 5.5 max 167.9Epoch's own ECI for GPT-6.1 Terra max: 167.7GPT-6.1 Terra max: 167.5, likely 165.3 to 169.7GPT-6.1 Terra max 167.5
0255075100120140160180Score on the benchmarkECI-style scoreTop model todayFable 5.2 max: 76.4 on AutomationBench-AAGPT-6.2 Astra max: 77.2 on AutomationBench-AAGemini 4 Neon xhigh: 79.3 on AutomationBench-AAGemini 4 Carbon high: 77.8 on AutomationBench-AAGPT-6.2 Sol max: 71.3 on AutomationBench-AAOpus 5.5 max: 69.5 on AutomationBench-AAGPT-6 Astra max: 68.5 on AutomationBench-AAGPT-6.1 Sol max: 64.9 on AutomationBench-AASonnet 5.5 max: 71.8 on AutomationBench-AAGPT-6.1 Terra max: 60.2 on AutomationBench-AAGemini 4 Hydrogen xhigh: 76.1 on AutomationBench-AAGemini 4 Argon high: 77.5 on AutomationBench-AAFable 5.1 xhigh: 57.8 on AutomationBench-AAMuse Wonder Full max: 71.3 on AutomationBench-AAKimi K3.1 max: 73.3 on AutomationBench-AAGPT-6 Sol max: 61.6 on AutomationBench-AAOpus 5 max: 56.6 on AutomationBench-AAMuse Wonder max: 68.1 on AutomationBench-AAQwen 4 Max xhigh: 72.6 on AutomationBench-AAGPT-5.6 Sol max: 60.1 on AutomationBench-AAHy4: 65.6 on AutomationBench-AAHaiku 5.5 max: 35.4 on AutomationBench-AAMuse Ember max: 68.1 on AutomationBench-AAMuse Spark 1.3 max: 57.9 on AutomationBench-AAGPT-5.6 Terra max: 59.6 on AutomationBench-AAGrok 4.7 xhigh: 65.6 on AutomationBench-AADeepSeek V4.1 Flash 1017 max: 66.6 on AutomationBench-AAQwen 3.8 Max xhigh: 56.2 on AutomationBench-AAGemini 3.8 Flash high: 59.9 on AutomationBench-AAGLM 5.3 max: 62.2 on AutomationBench-AADeepSeek V4.1 Flash max: 68.9 on AutomationBench-AAMiniMax M3.1 max: 72.5 on AutomationBench-AAGrok 4.6 high: 66.7 on AutomationBench-AAKimi K3 max: 58.3 on AutomationBench-AAMiMo V2.6 Pro: 58.6 on AutomationBench-AALing 3.1 Flash: 67.5 on AutomationBench-AAInkling 2 Small: 45.6 on AutomationBench-AAGPT-6 Luna max: 53.2 on AutomationBench-AAJoyAI-VL-Preview max: 61.7 on AutomationBench-AAStep 5 Preview: 51.0 on AutomationBench-AAQwen 3.8 Flash: 55.9 on AutomationBench-AAGLM 5.3 Flash max: 60.4 on AutomationBench-AAMiMo V2.6 Flash: 64.1 on AutomationBench-AAQwen 3.8 2.4T-A95B: 57.2 on AutomationBench-AAMistral Large 4 high: 57.8 on AutomationBench-AAGPT-5.6 Luna max: 50.2 on AutomationBench-AADeepSeek V4 Pro 0813 max: 56.7 on AutomationBench-AASonnet 5 max: 36.5 on AutomationBench-AAMuse Spark Open xhigh: 49.7 on AutomationBench-AAAgnes 2.5 Pro Theta: 48.8 on AutomationBench-AAMiniMax M3: 21.3 on AutomationBench-AAJT-4.1 Flash 236B A21B: 36.7 on AutomationBench-AAInkling 2: 41.8 on AutomationBench-AANanbeige4.3-33B: 53.1 on AutomationBench-AAQwen 3.8 27B xhigh: 48.2 on AutomationBench-AABeam 1.0: 48.5 on AutomationBench-AAGLM 5.2 max: 28.4 on AutomationBench-AAK2 Horizon 375B A23B high: 37.2 on AutomationBench-AAApodex 1.1: 28.2 on AutomationBench-AAK2 Horizon 32B: 29.4 on AutomationBench-AAK2 Horizon MoVA 36B A4B: 26.1 on AutomationBench-AASolar Pro 4: 13.6 on AutomationBench-AAInkling Small: 4.8 on AutomationBench-AAInkling xhigh: 5.0 on AutomationBench-AANemotron 3.5 Super VL high: 14.8 on AutomationBench-AASolar Mini 4: 22.3 on AutomationBench-AANemotron 3 Ultra max: 3.0 on AutomationBench-AAGemini 3.5 Flash Lite high: 25.0 on AutomationBench-AAK2 Horizon 7B: 20.9 on AutomationBench-AAA.X-K2: 2.5 on AutomationBench-AAK2 Horizon 3.7B: 14.7 on AutomationBench-AAK-EXAONE 2.0: 3.9 on AutomationBench-AALing 3.0 Tiny: 0.3 on AutomationBench-AAMuse Glimmer high: 6.8 on AutomationBench-AALFM3-1.7B: 7.0 on AutomationBench-AANemotron 3.5 Lightning: 0.8 on AutomationBench-AAGranite 4.2 30B: 1.5 on AutomationBench-AAGranite 4.2 8B: 2.1 on AutomationBench-AAGranite 4.2 3B: 0.4 on AutomationBench-AAFable 5.2 max: 73.3 on Terminal-Bench-Science-AAGPT-6.2 Astra max: 78.1 on Terminal-Bench-Science-AAGemini 4 Neon xhigh: 70.0 on Terminal-Bench-Science-AAGemini 4 Carbon high: 64.8 on Terminal-Bench-Science-AAGPT-6.2 Sol max: 67.6 on Terminal-Bench-Science-AAOpus 5.5 max: 59.0 on Terminal-Bench-Science-AAGPT-6 Astra max: 63.3 on Terminal-Bench-Science-AAGPT-6.1 Sol max: 58.1 on Terminal-Bench-Science-AASonnet 5.5 max: 53.3 on Terminal-Bench-Science-AAGPT-6.1 Terra max: 57.6 on Terminal-Bench-Science-AAGemini 4 Hydrogen xhigh: 49.5 on Terminal-Bench-Science-AAGemini 4 Argon high: 52.9 on Terminal-Bench-Science-AAFable 5.1 xhigh: 40.5 on Terminal-Bench-Science-AAMuse Wonder Full max: 37.1 on Terminal-Bench-Science-AAKimi K3.1 max: 34.3 on Terminal-Bench-Science-AAGPT-6 Sol max: 30.0 on Terminal-Bench-Science-AAOpus 5 max: 28.6 on Terminal-Bench-Science-AAMuse Wonder max: 25.7 on Terminal-Bench-Science-AAQwen 4 Max xhigh: 23.3 on Terminal-Bench-Science-AAGPT-5.6 Sol max: 22.4 on Terminal-Bench-Science-AAHy4: 12.9 on Terminal-Bench-Science-AAHaiku 5.5 max: 20.0 on Terminal-Bench-Science-AAMuse Ember max: 20.0 on Terminal-Bench-Science-AAMuse Spark 1.3 max: 11.0 on Terminal-Bench-Science-AAGPT-5.6 Terra max: 13.8 on Terminal-Bench-Science-AAGrok 4.7 xhigh: 11.4 on Terminal-Bench-Science-AADeepSeek V4.1 Flash 1017 max: 12.9 on Terminal-Bench-Science-AAQwen 3.8 Max xhigh: 11.9 on Terminal-Bench-Science-AAGemini 3.8 Flash high: 10.0 on Terminal-Bench-Science-AAGLM 5.3 max: 9.5 on Terminal-Bench-Science-AADeepSeek V4.1 Flash max: 9.0 on Terminal-Bench-Science-AAMiniMax M3.1 max: 3.3 on Terminal-Bench-Science-AAGrok 4.6 high: 6.2 on Terminal-Bench-Science-AAKimi K3 max: 8.1 on Terminal-Bench-Science-AAMiMo V2.6 Pro: 5.7 on Terminal-Bench-Science-AALing 3.1 Flash: 4.3 on Terminal-Bench-Science-AAInkling 2 Small: 5.7 on Terminal-Bench-Science-AAGPT-6 Luna max: 8.6 on Terminal-Bench-Science-AAJoyAI-VL-Preview max: 3.8 on Terminal-Bench-Science-AAStep 5 Preview: 2.4 on Terminal-Bench-Science-AAQwen 3.8 Flash: 7.1 on Terminal-Bench-Science-AAGLM 5.3 Flash max: 4.8 on Terminal-Bench-Science-AAMiMo V2.6 Flash: 4.3 on Terminal-Bench-Science-AAMistral Large 4 high: 4.3 on Terminal-Bench-Science-AAGPT-5.6 Luna max: 3.3 on Terminal-Bench-Science-AADeepSeek V4 Pro 0813 max: 5.7 on Terminal-Bench-Science-AAMuse Spark Open xhigh: 1.9 on Terminal-Bench-Science-AAAgnes 2.5 Pro Theta: 3.8 on Terminal-Bench-Science-AAMiniMax M3: 0.5 on Terminal-Bench-Science-AAJT-4.1 Flash 236B A21B: 0.0 on Terminal-Bench-Science-AAInkling 2: 1.4 on Terminal-Bench-Science-AANanbeige4.3-33B: 1.0 on Terminal-Bench-Science-AAQwen 3.8 27B xhigh: 0.5 on Terminal-Bench-Science-AABeam 1.0: 0.5 on Terminal-Bench-Science-AAK2 Horizon 375B A23B high: 0.0 on Terminal-Bench-Science-AAApodex 1.1: 0.5 on Terminal-Bench-Science-AAK2 Horizon 32B: 0.0 on Terminal-Bench-Science-AAK2 Horizon MoVA 36B A4B: 0.0 on Terminal-Bench-Science-AASolar Pro 4: 0.0 on Terminal-Bench-Science-AAInkling Small: 0.0 on Terminal-Bench-Science-AAInkling xhigh: 0.0 on Terminal-Bench-Science-AANemotron 3.5 Super VL high: 0.0 on Terminal-Bench-Science-AANemotron 3 Ultra max: 0.0 on Terminal-Bench-Science-AAGemini 3.5 Flash Lite high: 0.0 on Terminal-Bench-Science-AAK2 Horizon 7B: 0.0 on Terminal-Bench-Science-AAA.X-K2: 0.0 on Terminal-Bench-Science-AAK2 Horizon 3.7B: 0.0 on Terminal-Bench-Science-AAK-EXAONE 2.0: 0.0 on Terminal-Bench-Science-AAMuse Glimmer high: 0.0 on Terminal-Bench-Science-AALFM3-1.7B: 0.0 on Terminal-Bench-Science-AANemotron 3.5 Lightning: 0.0 on Terminal-Bench-Science-AAGranite 4.2 30B: 0.0 on Terminal-Bench-Science-AAGranite 4.2 8B: 0.0 on Terminal-Bench-Science-AAGranite 4.2 3B: 0.0 on Terminal-Bench-Science-AAFable 5.2 max: 24.2 on GDP.pdfGPT-6.2 Astra max: 29.8 on GDP.pdfGemini 4 Neon xhigh: 28.0 on GDP.pdfGemini 4 Carbon high: 28.4 on GDP.pdfGPT-6.2 Sol max: 30.0 on GDP.pdfOpus 5.5 max: 26.2 on GDP.pdfGPT-6 Astra max: 31.0 on GDP.pdfGPT-6.1 Sol max: 31.0 on GDP.pdfSonnet 5.5 max: 25.8 on GDP.pdfGPT-6.1 Terra max: 28.6 on GDP.pdfGemini 4 Hydrogen xhigh: 27.6 on GDP.pdfGemini 4 Argon high: 21.8 on GDP.pdfFable 5.1 xhigh: 26.2 on GDP.pdfMuse Wonder Full max: 32.6 on GDP.pdfKimi K3.1 max: 31.4 on GDP.pdfGPT-6 Sol max: 25.2 on GDP.pdfOpus 5 max: 21.6 on GDP.pdfMuse Wonder max: 30.6 on GDP.pdfQwen 4 Max xhigh: 26.2 on GDP.pdfGPT-5.6 Sol max: 27.2 on GDP.pdfHy4: 21.6 on GDP.pdfHaiku 5.5 max: 20.8 on GDP.pdfMuse Ember max: 22.8 on GDP.pdfMuse Spark 1.3 max: 26.6 on GDP.pdfGPT-5.6 Terra max: 24.0 on GDP.pdfGrok 4.7 xhigh: 20.0 on GDP.pdfDeepSeek V4.1 Flash 1017 max: 17.2 on GDP.pdfQwen 3.8 Max xhigh: 22.8 on GDP.pdfGemini 3.8 Flash high: 21.0 on GDP.pdfGLM 5.3 max: 11.2 on GDP.pdfDeepSeek V4.1 Flash max: 12.8 on GDP.pdfMiniMax M3.1 max: 20.8 on GDP.pdfGrok 4.6 high: 17.0 on GDP.pdfKimi K3 max: 22.0 on GDP.pdfMiMo V2.6 Pro: 19.2 on GDP.pdfLing 3.1 Flash: 12.6 on GDP.pdfInkling 2 Small: 25.2 on GDP.pdfGPT-6 Luna max: 22.8 on GDP.pdfJoyAI-VL-Preview max: 20.6 on GDP.pdfStep 5 Preview: 14.8 on GDP.pdfQwen 3.8 Flash: 15.6 on GDP.pdfGLM 5.3 Flash max: 15.4 on GDP.pdfMiMo V2.6 Flash: 9.0 on GDP.pdfQwen 3.8 2.4T-A95B: 15.6 on GDP.pdfMistral Large 4 high: 17.6 on GDP.pdfGPT-5.6 Luna max: 24.0 on GDP.pdfDeepSeek V4 Pro 0813 max: 11.4 on GDP.pdfSonnet 5 max: 13.2 on GDP.pdfMuse Spark Open xhigh: 18.6 on GDP.pdfAgnes 2.5 Pro Theta: 14.2 on GDP.pdfMiniMax M3: 9.8 on GDP.pdfJT-4.1 Flash 236B A21B: 6.6 on GDP.pdfInkling 2: 16.8 on GDP.pdfNanbeige4.3-33B: 11.2 on GDP.pdfQwen 3.8 27B xhigh: 16.4 on GDP.pdfBeam 1.0: 11.6 on GDP.pdfGLM 5.2 max: 10.4 on GDP.pdfK2 Horizon 375B A23B high: 7.4 on GDP.pdfApodex 1.1: 11.2 on GDP.pdfK2 Horizon 32B: 3.8 on GDP.pdfK2 Horizon MoVA 36B A4B: 4.2 on GDP.pdfSolar Pro 4: 7.0 on GDP.pdfInkling Small: 11.2 on GDP.pdfInkling xhigh: 12.8 on GDP.pdfNemotron 3.5 Super VL high: 8.0 on GDP.pdfSolar Mini 4: 5.0 on GDP.pdfNemotron 3 Ultra max: 5.0 on GDP.pdfGemini 3.5 Flash Lite high: 13.6 on GDP.pdfK2 Horizon 7B: 3.2 on GDP.pdfA.X-K2: 6.4 on GDP.pdfK2 Horizon 3.7B: 3.2 on GDP.pdfK-EXAONE 2.0: 4.2 on GDP.pdfLing 3.0 Tiny: 0.8 on GDP.pdfMuse Glimmer high: 10.0 on GDP.pdfLFM3-1.7B: 0.4 on GDP.pdfNemotron 3.5 Lightning: 3.0 on GDP.pdfGranite 4.2 30B: 1.8 on GDP.pdfGranite 4.2 8B: 0.6 on GDP.pdfGranite 4.2 3B: 0.2 on GDP.pdfTop models, with likely rangeEpoch's own ECI for Fable 5.2 max: 173.4Fable 5.2 max: 171.8, likely 169.1 to 174.5Fable 5.2 max 171.8Epoch's own ECI for GPT-6.2 Astra max: 172.8GPT-6.2 Astra max: 171.5, likely 168.9 to 174.1GPT-6.2 Astra max 171.5Gemini 4 Neon xhigh: 170.3, likely 167.9 to 172.8Gemini 4 Neon xhigh 170.3Gemini 4 Carbon high: 169.7, likely 167.4 to 172.1Gemini 4 Carbon high 169.7GPT-6.2 Sol max: 169.6, likely 167.2 to 171.9GPT-6.2 Sol max 169.6Epoch's own ECI for Opus 5.5 max: 168.7Opus 5.5 max: 169.3, likely 167.0 to 171.6Opus 5.5 max 169.3Epoch's own ECI for GPT-6 Astra max: 169.2GPT-6 Astra max: 169.2, likely 166.9 to 171.5GPT-6 Astra max 169.2Epoch's own ECI for GPT-6.1 Sol max: 168.6GPT-6.1 Sol max: 168.3, likely 166.1 to 170.5GPT-6.1 Sol max 168.3Epoch's own ECI for Sonnet 5.5 max: 166.6Sonnet 5.5 max: 167.9, likely 165.7 to 170.1Sonnet 5.5 max 167.9Epoch's own ECI for GPT-6.1 Terra max: 167.7GPT-6.1 Terra max: 167.5, likely 165.3 to 169.7GPT-6.1 Terra max 167.5

This method runs the same kind of fit on the benchmarks in your mix. It is not Epoch’s index, which Priors also carries, unchanged, as the benchmark Epoch Capabilities Index. It differs in these ways:

  • Benchmarks. Epoch fits over 50 benchmarks it chose. This uses only the ones in your mix, including Elo ratings converted to 0 to 100, which Epoch leaves out.
  • Weights. Epoch has none. Here a benchmark’s errors count in proportion to its weight, so a benchmark at 20% pulls twice as hard on capability as one at 10%.
  • Fit. Since November 4, 2026 Epoch samples a Bayesian version of the model, which takes minutes. This uses the method Epoch used until then, least squares with a small penalty on every parameter, because it has to run while you drag a slider.
  • Scale. Epoch pins Claude 3.5 Sonnet at 130 and GPT-5 at 150. Neither has results on today’s benchmarks, so the scale here is lined up with Epoch’s published scores instead: over the models both score, the average and the spread are made to match. If a ranking has fewer than five such models, the middle model reads 150.
  • Which models. Epoch scores a model with at least four benchmark results in the middle of the range. Priors keeps its own coverage rule, the same one the other methods use.
  • Ranges. Epoch reports 90% intervals from its samples. Priors shows a 95% range worked out from how well the fit explains each model’s scores.

When it differs. In the same direction as IRT ability: results on hard benchmarks count for more than the same number of points on easy ones. Unlike IRT it uses your mix alone, and it reports capability itself, so the numbers are not 0 to 100.

Limits. Epoch says of its own index that a score means little by itself and is for comparing models. That holds here too. A model with scores near zero on everything gets a very low number with a very wide range. With one benchmark in the mix it can only repeat that benchmark’s order.

Why one stops at 100 and the other does not

Both methods begin the same way: a fit that places every model at a position on one axis. They differ in what they report from that position.

  • IRT ability reports a predicted score. It reads the benchmark curves at the model’s position and averages them. Every curve flattens as it nears 100, so the score does too.
  • ECI-style reports the position itself. It puts the position through a straight line onto Epoch’s scale. A straight line never flattens, so the score has no ceiling.
IRT abilityGemini 4 Neon xhigh sits at 0.75. The curves read there, weighted: 64.2
ECI-styleIts own fit puts the same model at 0.628. On the line: 153.13 + 27.39 × 0.628 = 170.3
Move a model up the axisTop: the score the curves expect on this mix at each position, which is what IRT ability reports. Bottom: the straight line ECI-style uses. Drag to the right. The top reading closes in on 100 while the bottom one keeps climbing at the same rate.
0255075100IRT ability: the score the curves expect on the mixGemini 4 Neon xhigh: ability 0.75, IRT ability score 64.2Fable 5.2 max: ability 0.73, IRT ability score 62.8Gemini 4 Carbon high: ability 0.73, IRT ability score 62.7GPT-6.2 Astra max: ability 0.68, IRT ability score 60.2Opus 5.5 max: ability 0.65, IRT ability score 58.5Gemini 4 Hydrogen xhigh: ability 0.63, IRT ability score 57.3GPT-6 Astra max: ability 0.60, IRT ability score 56.1GPT-6.2 Sol max: ability 0.60, IRT ability score 55.8Sonnet 5.5 max: ability 0.58, IRT ability score 54.6GPT-6.1 Sol max: ability 0.56, IRT ability score 53.7Gemini 4 Argon high: ability 0.56, IRT ability score 53.4Muse Wonder Full max: ability 0.55, IRT ability score 53.1GPT-6.1 Terra max: ability 0.53, IRT ability score 51.7Fable 5.1 xhigh: ability 0.45, IRT ability score 47.4Kimi K3.1 max: ability 0.45, IRT ability score 47.0Muse Wonder max: ability 0.43, IRT ability score 45.9Qwen 4 Max xhigh: ability 0.41, IRT ability score 44.8Opus 5 max: ability 0.37, IRT ability score 42.8GPT-6 Sol max: ability 0.36, IRT ability score 42.3GPT-5.6 Sol max: ability 0.34, IRT ability score 41.3Muse Ember max: ability 0.29, IRT ability score 38.7Muse Spark 1.3 max: ability 0.29, IRT ability score 38.7Grok 4.7 high: ability 0.29, IRT ability score 38.6Haiku 5.5 max: ability 0.28, IRT ability score 38.0Qwen 3.8 Max xhigh: ability 0.23, IRT ability score 35.6Inkling 2 Small: ability 0.19, IRT ability score 33.7Grok 4.6 xhigh: ability 0.16, IRT ability score 32.1GLM 5.3 max: ability 0.15, IRT ability score 31.7DeepSeek V4.1 Flash 1017 max: ability 0.14, IRT ability score 31.4Hy4: ability 0.13, IRT ability score 30.8Ling 3.1 Flash: ability 0.11, IRT ability score 30.3Gemini 3.8 Flash high: ability 0.11, IRT ability score 30.1Step 5 Preview: ability 0.11, IRT ability score 29.9DeepSeek V4.1 Flash max: ability 0.11, IRT ability score 29.9Qwen 3.8 2.4T-A95B: ability 0.10, IRT ability score 29.8MiMo V2.6 Pro: ability 0.08, IRT ability score 29.0Qwen 3.8 Flash: ability 0.08, IRT ability score 28.9Kimi K3 max: ability 0.08, IRT ability score 28.8GLM 5.3 Flash max: ability 0.06, IRT ability score 28.1GPT-5.6 Terra max: ability 0.06, IRT ability score 28.0Mistral Large 4 high: ability 0.05, IRT ability score 27.6Muse Spark Open xhigh: ability 0.03, IRT ability score 27.0JoyAI-VL-Preview max: ability 0.02, IRT ability score 26.7MiMo V2.6 Flash: ability 0.01, IRT ability score 26.1Sonnet 5 max: ability 0.00, IRT ability score 25.9GPT-6 Luna max: ability -0.02, IRT ability score 25.3MiniMax M3.1 max: ability -0.02, IRT ability score 25.1Agnes 2.5 Pro Theta: ability -0.02, IRT ability score 25.0GLM 5.2 max: ability -0.09, IRT ability score 22.8Nanbeige4.3-33B: ability -0.09, IRT ability score 22.8Qwen 3.8 27B xhigh: ability -0.10, IRT ability score 22.3Inkling 2: ability -0.12, IRT ability score 21.9GPT-5.6 Luna max: ability -0.16, IRT ability score 20.5K2 Horizon 375B A23B high: ability -0.18, IRT ability score 20.0DeepSeek V4 Pro 0813 max: ability -0.21, IRT ability score 19.2Beam 1.0: ability -0.24, IRT ability score 18.6JT-4.1 Flash 236B A21B: ability -0.25, IRT ability score 18.1Solar Pro 4: ability -0.30, IRT ability score 17.1MiniMax M3: ability -0.35, IRT ability score 15.9K2 Horizon MoVA 36B A4B: ability -0.36, IRT ability score 15.7Nemotron 3.5 Super VL high: ability -0.37, IRT ability score 15.6K2 Horizon 32B: ability -0.37, IRT ability score 15.5Nemotron 3 Ultra max: ability -0.42, IRT ability score 14.5Solar Mini 4: ability -0.42, IRT ability score 14.5Gemini 3.5 Flash Lite high: ability -0.42, IRT ability score 14.4Inkling Small: ability -0.51, IRT ability score 12.9Apodex 1.1: ability -0.54, IRT ability score 12.3K2 Horizon 7B: ability -0.55, IRT ability score 12.1Inkling xhigh: ability -0.56, IRT ability score 12.0A.X-K2: ability -0.59, IRT ability score 11.5K-EXAONE 2.0: ability -0.69, IRT ability score 10.1LFM3-1.7B: ability -0.70, IRT ability score 10.0K2 Horizon 3.7B: ability -0.74, IRT ability score 9.5Granite 4.2 30B: ability -0.85, IRT ability score 8.2Nemotron 3.5 Lightning: ability -0.90, IRT ability score 7.7Granite 4.2 8B: ability -0.94, IRT ability score 7.4Granite 4.2 3B: ability -0.99, IRT ability score 7.0Muse Glimmer high: ability -1.12, IRT ability score 6.0Ling 3.0 Tiny: ability -1.36, IRT ability score 4.6North Mini Code: ability -1.60, IRT ability score 3.6K2 Horizon 0.9B: ability -1.75, IRT ability score 3.1LFM3-VL-2B: ability -2.02, IRT ability score 2.4Gemini 4 Neon xhigh100150200-2-1012ECI-style: a straight line of the positionPosition on the fitted axis (ability or capability; each method fits its own)Fable 5.2 max: capability 0.68, ECI-style score 171.8GPT-6.2 Astra max: capability 0.67, ECI-style score 171.5Gemini 4 Neon xhigh: capability 0.63, ECI-style score 170.3Gemini 4 Carbon high: capability 0.61, ECI-style score 169.7GPT-6.2 Sol max: capability 0.60, ECI-style score 169.6Opus 5.5 max: capability 0.59, ECI-style score 169.3GPT-6 Astra max: capability 0.59, ECI-style score 169.2GPT-6.1 Sol max: capability 0.55, ECI-style score 168.3Sonnet 5.5 max: capability 0.54, ECI-style score 167.9GPT-6.1 Terra max: capability 0.52, ECI-style score 167.5Gemini 4 Hydrogen xhigh: capability 0.52, ECI-style score 167.3Gemini 4 Argon high: capability 0.50, ECI-style score 166.8Fable 5.1 xhigh: capability 0.48, ECI-style score 166.4Muse Wonder Full max: capability 0.46, ECI-style score 165.7Kimi K3.1 max: capability 0.43, ECI-style score 165.0GPT-6 Sol max: capability 0.41, ECI-style score 164.4Opus 5 max: capability 0.40, ECI-style score 164.2Muse Wonder max: capability 0.38, ECI-style score 163.6Qwen 4 Max xhigh: capability 0.37, ECI-style score 163.3GPT-5.6 Sol max: capability 0.36, ECI-style score 163.0Hy4: capability 0.29, ECI-style score 161.0Haiku 5.5 max: capability 0.29, ECI-style score 161.0Muse Ember max: capability 0.28, ECI-style score 160.8Muse Spark 1.3 max: capability 0.27, ECI-style score 160.6GPT-5.6 Terra max: capability 0.27, ECI-style score 160.4Grok 4.7 xhigh: capability 0.24, ECI-style score 159.6DeepSeek V4.1 Flash 1017 max: capability 0.23, ECI-style score 159.5Qwen 3.8 Max xhigh: capability 0.23, ECI-style score 159.3Gemini 3.8 Flash high: capability 0.21, ECI-style score 159.0GLM 5.3 max: capability 0.21, ECI-style score 158.9DeepSeek V4.1 Flash max: capability 0.21, ECI-style score 158.7MiniMax M3.1 max: capability 0.19, ECI-style score 158.5Grok 4.6 high: capability 0.19, ECI-style score 158.4Kimi K3 max: capability 0.19, ECI-style score 158.3MiMo V2.6 Pro: capability 0.18, ECI-style score 158.1Ling 3.1 Flash: capability 0.17, ECI-style score 157.8Inkling 2 Small: capability 0.16, ECI-style score 157.5GPT-6 Luna max: capability 0.16, ECI-style score 157.4JoyAI-VL-Preview max: capability 0.16, ECI-style score 157.4Step 5 Preview: capability 0.15, ECI-style score 157.2Qwen 3.8 Flash: capability 0.13, ECI-style score 156.7GLM 5.3 Flash max: capability 0.13, ECI-style score 156.7MiMo V2.6 Flash: capability 0.11, ECI-style score 156.3Qwen 3.8 2.4T-A95B: capability 0.11, ECI-style score 156.1Mistral Large 4 high: capability 0.08, ECI-style score 155.4GPT-5.6 Luna max: capability 0.08, ECI-style score 155.3DeepSeek V4 Pro 0813 max: capability 0.07, ECI-style score 155.0Sonnet 5 max: capability 0.07, ECI-style score 155.0Muse Spark Open xhigh: capability 0.06, ECI-style score 154.7Agnes 2.5 Pro Theta: capability 0.03, ECI-style score 154.0MiniMax M3: capability -0.06, ECI-style score 151.6JT-4.1 Flash 236B A21B: capability -0.08, ECI-style score 151.0Inkling 2: capability -0.09, ECI-style score 150.6Nanbeige4.3-33B: capability -0.10, ECI-style score 150.5Qwen 3.8 27B xhigh: capability -0.12, ECI-style score 149.9Beam 1.0: capability -0.12, ECI-style score 149.8GLM 5.2 max: capability -0.14, ECI-style score 149.3K2 Horizon 375B A23B high: capability -0.18, ECI-style score 148.3Apodex 1.1: capability -0.26, ECI-style score 146.0K2 Horizon 32B: capability -0.31, ECI-style score 144.5K2 Horizon MoVA 36B A4B: capability -0.34, ECI-style score 143.9Solar Pro 4: capability -0.34, ECI-style score 143.7Inkling Small: capability -0.37, ECI-style score 143.1Inkling xhigh: capability -0.37, ECI-style score 143.0Nemotron 3.5 Super VL high: capability -0.37, ECI-style score 142.9Solar Mini 4: capability -0.49, ECI-style score 139.8Nemotron 3 Ultra max: capability -0.51, ECI-style score 139.2Gemini 3.5 Flash Lite high: capability -0.52, ECI-style score 139.0K2 Horizon 7B: capability -0.52, ECI-style score 138.9A.X-K2: capability -0.73, ECI-style score 133.2K2 Horizon 3.7B: capability -0.73, ECI-style score 133.0K-EXAONE 2.0: capability -0.88, ECI-style score 129.2Ling 3.0 Tiny: capability -0.93, ECI-style score 127.7Muse Glimmer high: capability -0.95, ECI-style score 127.1LFM3-1.7B: capability -1.13, ECI-style score 122.2Nemotron 3.5 Lightning: capability -1.14, ECI-style score 121.8Granite 4.2 30B: capability -1.17, ECI-style score 121.1Granite 4.2 8B: capability -1.20, ECI-style score 120.2Granite 4.2 3B: capability -1.26, ECI-style score 118.6Fable 5.2 max64.2173.8
IRT ability score 64.2one more unit adds 26.9ECI-style score 173.8one more unit adds 27.4, alwaysPosition 0.75
0255075100IRT ability: the score the curves expect on the mixGemini 4 Neon xhigh: ability 0.75, IRT ability score 64.2Fable 5.2 max: ability 0.73, IRT ability score 62.8Gemini 4 Carbon high: ability 0.73, IRT ability score 62.7GPT-6.2 Astra max: ability 0.68, IRT ability score 60.2Opus 5.5 max: ability 0.65, IRT ability score 58.5Gemini 4 Hydrogen xhigh: ability 0.63, IRT ability score 57.3GPT-6 Astra max: ability 0.60, IRT ability score 56.1GPT-6.2 Sol max: ability 0.60, IRT ability score 55.8Sonnet 5.5 max: ability 0.58, IRT ability score 54.6GPT-6.1 Sol max: ability 0.56, IRT ability score 53.7Gemini 4 Argon high: ability 0.56, IRT ability score 53.4Muse Wonder Full max: ability 0.55, IRT ability score 53.1GPT-6.1 Terra max: ability 0.53, IRT ability score 51.7Fable 5.1 xhigh: ability 0.45, IRT ability score 47.4Kimi K3.1 max: ability 0.45, IRT ability score 47.0Muse Wonder max: ability 0.43, IRT ability score 45.9Qwen 4 Max xhigh: ability 0.41, IRT ability score 44.8Opus 5 max: ability 0.37, IRT ability score 42.8GPT-6 Sol max: ability 0.36, IRT ability score 42.3GPT-5.6 Sol max: ability 0.34, IRT ability score 41.3Muse Ember max: ability 0.29, IRT ability score 38.7Muse Spark 1.3 max: ability 0.29, IRT ability score 38.7Grok 4.7 high: ability 0.29, IRT ability score 38.6Haiku 5.5 max: ability 0.28, IRT ability score 38.0Qwen 3.8 Max xhigh: ability 0.23, IRT ability score 35.6Inkling 2 Small: ability 0.19, IRT ability score 33.7Grok 4.6 xhigh: ability 0.16, IRT ability score 32.1GLM 5.3 max: ability 0.15, IRT ability score 31.7DeepSeek V4.1 Flash 1017 max: ability 0.14, IRT ability score 31.4Hy4: ability 0.13, IRT ability score 30.8Ling 3.1 Flash: ability 0.11, IRT ability score 30.3Gemini 3.8 Flash high: ability 0.11, IRT ability score 30.1Step 5 Preview: ability 0.11, IRT ability score 29.9DeepSeek V4.1 Flash max: ability 0.11, IRT ability score 29.9Qwen 3.8 2.4T-A95B: ability 0.10, IRT ability score 29.8MiMo V2.6 Pro: ability 0.08, IRT ability score 29.0Qwen 3.8 Flash: ability 0.08, IRT ability score 28.9Kimi K3 max: ability 0.08, IRT ability score 28.8GLM 5.3 Flash max: ability 0.06, IRT ability score 28.1GPT-5.6 Terra max: ability 0.06, IRT ability score 28.0Mistral Large 4 high: ability 0.05, IRT ability score 27.6Muse Spark Open xhigh: ability 0.03, IRT ability score 27.0JoyAI-VL-Preview max: ability 0.02, IRT ability score 26.7MiMo V2.6 Flash: ability 0.01, IRT ability score 26.1Sonnet 5 max: ability 0.00, IRT ability score 25.9GPT-6 Luna max: ability -0.02, IRT ability score 25.3MiniMax M3.1 max: ability -0.02, IRT ability score 25.1Agnes 2.5 Pro Theta: ability -0.02, IRT ability score 25.0GLM 5.2 max: ability -0.09, IRT ability score 22.8Nanbeige4.3-33B: ability -0.09, IRT ability score 22.8Qwen 3.8 27B xhigh: ability -0.10, IRT ability score 22.3Inkling 2: ability -0.12, IRT ability score 21.9GPT-5.6 Luna max: ability -0.16, IRT ability score 20.5K2 Horizon 375B A23B high: ability -0.18, IRT ability score 20.0DeepSeek V4 Pro 0813 max: ability -0.21, IRT ability score 19.2Beam 1.0: ability -0.24, IRT ability score 18.6JT-4.1 Flash 236B A21B: ability -0.25, IRT ability score 18.1Solar Pro 4: ability -0.30, IRT ability score 17.1MiniMax M3: ability -0.35, IRT ability score 15.9K2 Horizon MoVA 36B A4B: ability -0.36, IRT ability score 15.7Nemotron 3.5 Super VL high: ability -0.37, IRT ability score 15.6K2 Horizon 32B: ability -0.37, IRT ability score 15.5Nemotron 3 Ultra max: ability -0.42, IRT ability score 14.5Solar Mini 4: ability -0.42, IRT ability score 14.5Gemini 3.5 Flash Lite high: ability -0.42, IRT ability score 14.4Inkling Small: ability -0.51, IRT ability score 12.9Apodex 1.1: ability -0.54, IRT ability score 12.3K2 Horizon 7B: ability -0.55, IRT ability score 12.1Inkling xhigh: ability -0.56, IRT ability score 12.0A.X-K2: ability -0.59, IRT ability score 11.5K-EXAONE 2.0: ability -0.69, IRT ability score 10.1LFM3-1.7B: ability -0.70, IRT ability score 10.0K2 Horizon 3.7B: ability -0.74, IRT ability score 9.5Granite 4.2 30B: ability -0.85, IRT ability score 8.2Nemotron 3.5 Lightning: ability -0.90, IRT ability score 7.7Granite 4.2 8B: ability -0.94, IRT ability score 7.4Granite 4.2 3B: ability -0.99, IRT ability score 7.0Muse Glimmer high: ability -1.12, IRT ability score 6.0Ling 3.0 Tiny: ability -1.36, IRT ability score 4.6North Mini Code: ability -1.60, IRT ability score 3.6K2 Horizon 0.9B: ability -1.75, IRT ability score 3.1LFM3-VL-2B: ability -2.02, IRT ability score 2.4Gemini 4 Neon xhigh100150200-2-1012ECI-style: a straight line of the positionPosition on the fitted axisFable 5.2 max: capability 0.68, ECI-style score 171.8GPT-6.2 Astra max: capability 0.67, ECI-style score 171.5Gemini 4 Neon xhigh: capability 0.63, ECI-style score 170.3Gemini 4 Carbon high: capability 0.61, ECI-style score 169.7GPT-6.2 Sol max: capability 0.60, ECI-style score 169.6Opus 5.5 max: capability 0.59, ECI-style score 169.3GPT-6 Astra max: capability 0.59, ECI-style score 169.2GPT-6.1 Sol max: capability 0.55, ECI-style score 168.3Sonnet 5.5 max: capability 0.54, ECI-style score 167.9GPT-6.1 Terra max: capability 0.52, ECI-style score 167.5Gemini 4 Hydrogen xhigh: capability 0.52, ECI-style score 167.3Gemini 4 Argon high: capability 0.50, ECI-style score 166.8Fable 5.1 xhigh: capability 0.48, ECI-style score 166.4Muse Wonder Full max: capability 0.46, ECI-style score 165.7Kimi K3.1 max: capability 0.43, ECI-style score 165.0GPT-6 Sol max: capability 0.41, ECI-style score 164.4Opus 5 max: capability 0.40, ECI-style score 164.2Muse Wonder max: capability 0.38, ECI-style score 163.6Qwen 4 Max xhigh: capability 0.37, ECI-style score 163.3GPT-5.6 Sol max: capability 0.36, ECI-style score 163.0Hy4: capability 0.29, ECI-style score 161.0Haiku 5.5 max: capability 0.29, ECI-style score 161.0Muse Ember max: capability 0.28, ECI-style score 160.8Muse Spark 1.3 max: capability 0.27, ECI-style score 160.6GPT-5.6 Terra max: capability 0.27, ECI-style score 160.4Grok 4.7 xhigh: capability 0.24, ECI-style score 159.6DeepSeek V4.1 Flash 1017 max: capability 0.23, ECI-style score 159.5Qwen 3.8 Max xhigh: capability 0.23, ECI-style score 159.3Gemini 3.8 Flash high: capability 0.21, ECI-style score 159.0GLM 5.3 max: capability 0.21, ECI-style score 158.9DeepSeek V4.1 Flash max: capability 0.21, ECI-style score 158.7MiniMax M3.1 max: capability 0.19, ECI-style score 158.5Grok 4.6 high: capability 0.19, ECI-style score 158.4Kimi K3 max: capability 0.19, ECI-style score 158.3MiMo V2.6 Pro: capability 0.18, ECI-style score 158.1Ling 3.1 Flash: capability 0.17, ECI-style score 157.8Inkling 2 Small: capability 0.16, ECI-style score 157.5GPT-6 Luna max: capability 0.16, ECI-style score 157.4JoyAI-VL-Preview max: capability 0.16, ECI-style score 157.4Step 5 Preview: capability 0.15, ECI-style score 157.2Qwen 3.8 Flash: capability 0.13, ECI-style score 156.7GLM 5.3 Flash max: capability 0.13, ECI-style score 156.7MiMo V2.6 Flash: capability 0.11, ECI-style score 156.3Qwen 3.8 2.4T-A95B: capability 0.11, ECI-style score 156.1Mistral Large 4 high: capability 0.08, ECI-style score 155.4GPT-5.6 Luna max: capability 0.08, ECI-style score 155.3DeepSeek V4 Pro 0813 max: capability 0.07, ECI-style score 155.0Sonnet 5 max: capability 0.07, ECI-style score 155.0Muse Spark Open xhigh: capability 0.06, ECI-style score 154.7Agnes 2.5 Pro Theta: capability 0.03, ECI-style score 154.0MiniMax M3: capability -0.06, ECI-style score 151.6JT-4.1 Flash 236B A21B: capability -0.08, ECI-style score 151.0Inkling 2: capability -0.09, ECI-style score 150.6Nanbeige4.3-33B: capability -0.10, ECI-style score 150.5Qwen 3.8 27B xhigh: capability -0.12, ECI-style score 149.9Beam 1.0: capability -0.12, ECI-style score 149.8GLM 5.2 max: capability -0.14, ECI-style score 149.3K2 Horizon 375B A23B high: capability -0.18, ECI-style score 148.3Apodex 1.1: capability -0.26, ECI-style score 146.0K2 Horizon 32B: capability -0.31, ECI-style score 144.5K2 Horizon MoVA 36B A4B: capability -0.34, ECI-style score 143.9Solar Pro 4: capability -0.34, ECI-style score 143.7Inkling Small: capability -0.37, ECI-style score 143.1Inkling xhigh: capability -0.37, ECI-style score 143.0Nemotron 3.5 Super VL high: capability -0.37, ECI-style score 142.9Solar Mini 4: capability -0.49, ECI-style score 139.8Nemotron 3 Ultra max: capability -0.51, ECI-style score 139.2Gemini 3.5 Flash Lite high: capability -0.52, ECI-style score 139.0K2 Horizon 7B: capability -0.52, ECI-style score 138.9A.X-K2: capability -0.73, ECI-style score 133.2K2 Horizon 3.7B: capability -0.73, ECI-style score 133.0K-EXAONE 2.0: capability -0.88, ECI-style score 129.2Ling 3.0 Tiny: capability -0.93, ECI-style score 127.7Muse Glimmer high: capability -0.95, ECI-style score 127.1LFM3-1.7B: capability -1.13, ECI-style score 122.2Nemotron 3.5 Lightning: capability -1.14, ECI-style score 121.8Granite 4.2 30B: capability -1.17, ECI-style score 121.1Granite 4.2 8B: capability -1.20, ECI-style score 120.2Granite 4.2 3B: capability -1.26, ECI-style score 118.6Fable 5.2 max64.2173.8
IRT ability score 64.2one more unit adds 26.9ECI-style score 173.8one more unit adds 27.4, alwaysPosition 0.75

So the same gap between two strong models shows in full on the ECI-style scale and shrinks on the IRT ability scale as the mix gets closer to solved. Neither is more correct. A 0 to 100 score says how much of this mix a model is expected to get. An open scale keeps separating models after that, though with wider ranges, because a nearly solved benchmark says little about how much further a model could go.

The same mix, three ways

The top 10 on the General frontier mix as of 2026-11-22, each model at its best reasoning effort, with its likely range underneath.

#Weighted averageIRT abilityECI-style
1Fable 5.2 max 64.5Gemini 4 Neon xhigh 64.256.1 to 71.0Fable 5.2 max 171.8169.1 to 174.5
2GPT-6.2 Astra max 62.1Fable 5.2 max 62.854.8 to 69.8GPT-6.2 Astra max 171.5168.9 to 174.1
3Gemini 4 Neon xhigh 61.4Gemini 4 Carbon high 62.754.7 to 69.7Gemini 4 Neon xhigh 170.3167.9 to 172.8
4Gemini 4 Carbon high 60.0GPT-6.2 Astra max 60.252.1 to 67.4Gemini 4 Carbon high 169.7167.4 to 172.1
5Opus 5.5 max 58.9Opus 5.5 max 58.550.5 to 65.9GPT-6.2 Sol max 169.6167.2 to 171.9
6GPT-6.2 Sol max 57.3Gemini 4 Hydrogen xhigh 57.349.3 to 64.7Opus 5.5 max 169.3167.0 to 171.6
7GPT-6 Astra max 56.9GPT-6 Astra max 56.148.1 to 63.6GPT-6 Astra max 169.2166.9 to 171.5
8Sonnet 5.5 max 54.6GPT-6.2 Sol max 55.847.9 to 63.4GPT-6.1 Sol max 168.3166.1 to 170.5
9GPT-6.1 Sol max 54.6Sonnet 5.5 max 54.646.7 to 62.2Sonnet 5.5 max 167.9165.7 to 170.1
10Gemini 4 Hydrogen xhigh 53.9GPT-6.1 Sol max 53.745.8 to 61.3GPT-6.1 Terra max 167.5165.3 to 169.7

Open this mix with IRT ability or ECI-style scores.

What stays the same

  • Benchmark scores are used as published under every method.
  • Cost, speed and time per task do not depend on the method.
  • Rankings through time always use the weighted average, so one day compares with the next.
  • The method travels with the ranking: in its link, on a saved page and on its share card.