priors
ExplorePrivacySupport

Settings

Theme

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

Mode

01 Your calls02 Your ranking03 The evidence

Which is better?

Tap the better model, 10 quick calls at most. We find the benchmarks that agree with you. About a minute · no sign-in · 899 rankings so far

Skip to the data
Your top 5Call 01 · up to 10

Your gut calls, backed by benchmarks.

About a minute. No sign-in. · Privacy

Better evidence, fewer benchmarks.

Your preferences guide the search. Quality helps break close calls.

Coverage counts how many models have measured results — a test most models took beats a niche one.

Recency rewards recent test releases. An unknown release date gets a neutral score. A fresh scrape is not a fresh benchmark.

Headroom measures saturation and score spread. Tests that separate models beat tests everyone passes.

Objectivity favors verifiable outcomes over preferences and subjective judging.

Quality combines 25% coverage, 25% recency, 30% headroom, and 20% objectivity — plus a bonus for preferred benchmarks, minus a penalty for superseded versions. Objectivity is an assigned judgment score. This heuristic does not measure contamination or scientific validity.

BenchmarkRecentHeadroomObjectiveQuality
✳ ProgramBench v1 Fully Resolved877798100
Released 2026-05-05 · Why this score

Released 2026-05-05. Headroom and score-spread quality 77/100 after collapsing effort/config variants by model. Objectivity 98/100 is an assigned judgment score based on grading method. Breadth 100/100 from 62 measured models; frontier coverage 70/100 from 7/10 current top-lineup models. Publisher results updated 2026-10-12. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗

✳ FrontierMath Tier 4 (v2)93409690
Released 2026-06-12 · Why this score

Released 2026-06-12. Headroom and score-spread quality 40/100 after collapsing effort/config variants by model. Objectivity 96/100 is an assigned judgment score based on grading method. Breadth 100/100 from 67 measured models; frontier coverage 70/100 from 7/10 current top-lineup models. Publisher update date unknown. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗

✳ AA-Omniscience60498589
Released 2025-11-16 · Why this score

Released 2025-11-16. Headroom and score-spread quality 49/100 after collapsing effort/config variants by model. Objectivity 85/100 is an assigned judgment score based on grading method. Breadth 100/100 from 368 measured models; frontier coverage 100/100 from 10/10 current top-lineup models. Publisher update date unknown. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗

✳ Terminal-Bench 4.0100449284
Released 2026-08-01 · Why this score

Released 2026-08-01. Headroom and score-spread quality 44/100 after collapsing effort/config variants by model. Objectivity 92/100 is an assigned judgment score based on grading method. Breadth 57/100 from 23 measured models; frontier coverage 60/100 from 6/10 current top-lineup models. Publisher update date unknown. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗

✳ Humanity's Last Exam14608883
Released 2025-01-23 · Why this score

Released 2025-01-23. Headroom and score-spread quality 60/100 after collapsing effort/config variants by model. Objectivity 88/100 is an assigned judgment score based on grading method. Breadth 100/100 from 423 measured models; frontier coverage 100/100 from 10/10 current top-lineup models. Publisher update date unknown. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗

✳ BullshitBench v277416581
Released 2026-03-02 · Why this score

Released 2026-03-02. Headroom and score-spread quality 41/100 after collapsing effort/config variants by model. Objectivity 65/100 is an assigned judgment score based on grading method. Breadth 100/100 from 108 measured models; frontier coverage 70/100 from 7/10 current top-lineup models. Publisher update date unknown. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗

✳ ARC-AGI-381469079
Released 2026-03-25 · Why this score

Released 2026-03-25. Headroom and score-spread quality 46/100 after collapsing effort/config variants by model. Objectivity 90/100 is an assigned judgment score based on grading method. Breadth 57/100 from 23 measured models; frontier coverage 50/100 from 5/10 current top-lineup models. Publisher update date unknown. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗

✳ DeepSWE v1.193379677
Released 2026-06-14 · Why this score

Released 2026-06-14. Headroom and score-spread quality 37/100 after collapsing effort/config variants by model. Objectivity 96/100 is an assigned judgment score based on grading method. Breadth 70/100 from 28 measured models; frontier coverage 20/100 from 2/10 current top-lineup models. Publisher update date unknown. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗

✳ SimpleBench5378668
Released 2024-11-25 · Why this score

Released 2024-11-25. Headroom and score-spread quality 37/100 after collapsing effort/config variants by model. Objectivity 86/100 is an assigned judgment score based on grading method. Breadth 100/100 from 96 measured models; frontier coverage 60/100 from 6/10 current top-lineup models. Publisher update date unknown. Quality combines 15% breadth, 15% frontier coverage, 20% benchmark recency, 30% headroom and spread, and 20% objectivity, plus a preferred-benchmark bonus. It does not prove validity or freedom from training contamination. Source ↗