Support

Settings

Theme

Explore themes

Mode

Library

Looks borrowed from benchmark sites and products. Themes with a light and a dark version follow your mode.

Bring your own theme

Upload a Priors theme file, or have your agent make one: copy the prompt, tell it the look you want, and upload what it returns. With the Priors skill installed it already knows the format.

Custom themes stay in this browser. They travel inside shared chart links and a page's theme, so others see them as you do.

For the builder, Explore and model pages on this device. Published pages keep the theme their owner picked.

Mode

Update notices

A note when a benchmark one of your indexes uses gets a new version, with a one-click upgrade.

Other · released 2026-09-03

K2 Horizon 32B

Measured on 19 benchmarks.

Compare

13.9

#59 of 81 models

Strengths

  • K2 Horizon 32B
  • Fable 5.2 (#1)
  • Field median
bestField medianFrontierCode 2.0-AATerminal-Benchv4-AAAutomationBench-AAAA-Briefcase v1.1GDP.pdfTerminal-Bench-Science-AACritPt v2 HardSubsetAA-Omnisciencev1.1 AccuracyHLE-DiamondAA-Omnisciencev1.1Non-hallucination
  1. FrontierCode 2.0-AA15%2.3%#66 of 81
  2. Terminal-Bench v4-AA10%0.5%#63 of 81
  3. AutomationBench-AA5%29.4%#56 of 81
  4. AA-Briefcase v1.115%1175#54 of 81
  5. GDP.pdf5%3.8%#70 of 81
  6. Terminal-Bench-Science-AA15%0.0%#58 of 81
  7. CritPt v2 Hard Subset10%10.0%#63 of 81
  8. AA-Omniscience v1.1 Accuracy10%16.5%#69 of 81
  9. HLE-Diamond10%8.0%#67 of 81
  10. AA-Omniscience v1.1 Non-hallucination5%66.5%#31 of 81

Bars show where K2 Horizon 32B sits among the 81 ranked models on each benchmark (right is best); the tick is the field median.

By category

  • K2 Horizon 32B
  • Fable 5.2 (#1)
  • Field median
p25p50p75topField medianCodingAgents & computeruseProfessional workScience & researchReasoning &puzzlesKnowledge &accuracy
  1. Coding2 benchmarks20%#58 of 81
  2. Agents & computer use5 benchmarks26%#56 of 81
  3. Professional work3 benchmarks26%#57 of 81
  4. Science & research3 benchmarks20%#64 of 80
  5. Reasoning & puzzles2 benchmarks17%#62 of 79
  6. Knowledge & accuracy2 benchmarks18%#68 of 80

Across 17 results on every current benchmark that measured K2 Horizon 32B, not only this ranking’s. The percentage is its average standing among the 81 models in this ranking that took the same benchmarks: 100% beat them all, 0% trailed them all. A benchmark in two categories counts in both; a category needs two benchmarks.

Cost and speed

against 81 ranked models

Output tokens per task

#27 fewest of 73

74k

Not measured for K2 Horizon 32B: cost per task, time per task, output speed, time to first token (no published API price yet).

This ranking’s benchmarks

10 of 10 measured

Cost per task is what a subscriber pays: the model’s plan price, a share of API price. Models no plan sells are at API price.

Agents & computer use

AutomationBench-AA

282 results
  1. #166default of 28229.4%

Leader: Gemini 4 Neon · 79.3%

Professional work

AA-Briefcase v1.1

287 results
  1. #139default of 2871175

Leader: Sonnet 5.5 · 1815

Science & research

CritPt v2 Hard Subset

159 results
  1. #126default of 15910.0%

Leader: Fable 5.2 · 76.5%

Coding

FrontierCode 2.0-AA

156 results
  1. #131default of 1562.3%

Leader: Fable 5.2 · 49.1%

Professional work

GDP.pdf

281 results
  1. #227default of 2813.8%

Leader: Gemini 4 Neon · 33.4%

Knowledge & accuracy

HLE-Diamond

196 results
  1. #148default of 1968.0%

Leader: Fable 5.2 · 80.2%

Knowledge & accuracy

AA-Omniscience v1.1 Accuracy

169 results
  1. #152default of 16916.5%

Leader: Fable 5.2 · 79.1%

Knowledge & accuracy

AA-Omniscience v1.1 Non-hallucination

169 results
  1. #49default of 16966.5%

Leader: Gemini 4 Carbon · 95.8%

Science & research

Terminal-Bench-Science-AA

167 results
  1. #134default of 1670.0%

Leader: GPT-6.2 Astra · 78.6%

Agents & computer use

Terminal-Bench v4-AA

278 results
  1. #189default of 2780.5%

Leader: Kimi K3.1 · 67.2%

All 19 benchmarks

show 9 more