Atlas

Compare Models

Opus 4.8 vs GPT-5.6 Sol Pro

GPT-5.6 Sol Pro leads by 3.8 capability points (≈95% interval +1.5 to +6.2): a clear lead over Opus 4.8 · P(higher) ≈ 100%

Wins 16 of 23 shared benchmarks (5 lost, 2 tied)

Capability vs. relative cost
Opus 4.8GPT-5.6 Sol ProClosed weightsOpen weights

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting. GPT-5.6 Sol Pro has no setting with a relative cost yet.

Opus 4.8vsGPT-5.6 Sol Procapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins5 – 16 · 23 shared, 2 tied

Comparing matching setups only (83 benchmarks excluded) · suites counted once
← Opus 4.8GPT-5.6 Sol Pro →

Opus 4.8vsGPT-5.6 Sol Prowins by categoryData Nov 15, 2026Data: Mercor, ForecastBench, MineBench Maintainers and 9 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-5.6 Sol Pro minus Opus 4.8.

FrontierMath Tier 4 (v2) - PrivateMath56.180.5+24.4
BrowseComp (Mercor 130-question subset)Search73.795.7+22.0
MineBench Conservative RatingGames16862000+18.6
GeneBench-ProScience16.031.5+15.5
SWE Atlas - Codebase QnACode44.430.9−13.5
MineBench — Bradley-Terry RatingGames17861994+12.6
APEX-1 (Mercor, mean score)Professional Work62.773.1+10.4
BuseyBench SVGMultimodal5.46.3+9.6
Terminal-Bench 4.0Code23.632.3+8.7
Harvey LAB (Mercor run, 1,749 public tasks)Legal9.42.7−6.7
WeirdML (v2)Science82.989.4+6.5
RedlineBenchLegal39.745.8+6.1
HealthBench Professional (Length-Adjusted)Health & Medicine55.861.4+5.6
APEX-AccountingFinance & Accounting48.051.5+3.5
GPQA DiamondGeneral QA91.994.9+3.0
Ruby LLM Benchmarks — Program FixerCode72.675.3+2.7
Terminal-Bench 2.1Code82.081.2−0.8
SWE Atlas Codebase QnA Extended (Mercor held-out)Code24.024.7+0.7
APEX-AgentsProfessional Work39.440.0+0.6
ForecastBench Preliminary Dataset Brier IndexForecasting61.661.1−0.5
ForecastBench Tournament Overall Brier IndexForecasting66.265.9−0.3
SWE-bench Verified Extended (Mercor held-out)Code63.363.30.0
WordleBenchGames94.094.00.0

Opus 4.8vsGPT-5.6 Sol Pronormalised score marginsData Nov 15, 2026Data: Mercor, ForecastBench, MineBench Maintainers and 9 othersVillage LLM Atlasatlas.village.dev

Win counts use each benchmark’s declared score direction. Represented suite slices and the derived Capability score are excluded from headline totals. Missing or incompatible setups are excluded, never counted as losses.
Loading Atlas data…