Atlas

Compare Models

GPT-5.6 Sol Pro vs GPT-5.6 Terra

GPT-5.6 Sol Pro leads by 2.9 capability points (≈95% interval +0.5 to +5.3): a clear lead over GPT-5.6 Terra · P(higher) ≈ 99%

Wins 13 of 21 shared benchmarks (7 lost, 1 tied)

Capability vs. relative cost
GPT-5.6 Sol ProGPT-5.6 TerraClosed 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.

GPT-5.6 Sol ProvsGPT-5.6 Terracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins13 – 7 · 21 shared, 1 tied

Comparing matching setups only (83 benchmarks excluded) · suites counted once
← GPT-5.6 Sol ProGPT-5.6 Terra →

GPT-5.6 Sol ProvsGPT-5.6 Terrawins by categoryData Nov 15, 2026Data: Mercor, Epoch AI, Eval Connections and 7 othersVillage LLM Atlasatlas.village.dev

Margins

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

NanoGPT Speedrun Frontier (human record gap closed)Agents33.611.0−30.4
SWE-bench Verified Extended (Mercor held-out)Code63.378.7+15.4
SWE Atlas - Codebase QnACode30.945.2+14.3
APEX-AccountingFinance & Accounting51.540.3−11.2
WeirdML (v2)Science89.478.3−11.2
BrowseComp (Mercor 130-question subset)Search95.785.8−9.9
FrontierMath Tier 4 (v2) - PrivateMath80.570.7−9.8
SWE Atlas Codebase QnA Extended (Mercor held-out)Code24.716.0−8.7
GeneBench-ProScience31.523.3−8.2
Terminal-Bench 2.1Code81.288.0+6.8
APEX-Agents 1.1Professional Work65.271.8+6.6
Terminal-Bench 4.0Code32.326.3−6.0
GPQA DiamondGeneral QA94.989.9−5.0
SWE-bench VerifiedCode72.877.8+5.0
τ³-Banking (Mercor run, bm25_grep retrieval)Agents43.038.8−4.2
APEX-1 (Mercor, mean score)Professional Work73.169.5−3.6
WordleBenchGames94.091.0−3.0
RedlineBenchLegal45.847.8+2.0
Ruby LLM Benchmarks — Program FixerCode75.374.0−1.3
SWE-bench MultilingualCode94.294.4+0.2
Eval ConnectionsGames100.0100.00.0

GPT-5.6 Sol ProvsGPT-5.6 Terranormalised score marginsData Nov 15, 2026Data: Mercor, Epoch AI, Eval Connections and 7 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…