GPT-5.6 Sol vs GPT-5.6 Terra
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
Wins 0 of 8 areas
—
GPT-5.6 Sol is the stronger all-rounder.GPT-5.6 Terra is cheaper.
Both rank among the ten best models we track in coding.
Scores updated · 61 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking70GPT-5.6 Sol7 of 7 tests
- CodingWriting and fixing software60GPT-5.6 Sol6 of 7 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own60GPT-5.6 Sol6 of 6 tests
- MathCompetition and research-level math30GPT-5.6 Sol3 of 3 tests
- Following instructionsDoing exactly what it is asked20GPT-5.6 Sol2 of 2 tests
- FactsGetting facts right instead of making them up21GPT-5.6 Sol2 of 3 tests
- Images and chartsUnderstanding pictures, charts and video10GPT-5.6 Sol1 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts10GPT-5.6 Sol1 of 1 test
Long documents rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
GPT-5.6 Terra costs 42% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where GPT-5.6 Sol pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+26.5points ahead
- Common-sense trick questionsSimpleBench+15.9points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+14.7points ahead
Where GPT-5.6 Terra pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+4.3points ahead
Every test, side by side
All 61 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.6 Sol
- LMArena · WebDevGPT-5.6 Sol by 96 rating points16181522+96 rating
- Terminal-Bench HardGPT-5.6 Sol by 8.365.957.6+8.3
- LiveBench · CodingGPT-5.6 Sol by 5.683.978.3+5.6
- SciCodeGPT-5.6 Sol by 2.157.155+2.1
- LiveBench · Agentic CodingGPT-5.6 Sol by 1.256.255+1.2
- SWE-bench ProGPT-5.6 Sol by 1.264.663.4+1.2
- Terminal-Bench 2.1tie8888tie
AgentsGPT-5.6 Sol
- GDPValGPT-5.6 Sol by 7.955.647.7+7.9
- AA IT-Bench SREGPT-5.6 Sol by 5.256.251+5.2
- Terminal-Bench 4.0GPT-5.6 Sol by 4.539.935.4+4.5
- τ-Bench V3 · BankingGPT-5.6 Sol by 4.144.340.2+4.1
- BrowseCompGPT-5.6 Sol by 2.990.487.5+2.9
- AA ApexAgentsGPT-5.6 Sol by 139.938.9+1
ReasoningGPT-5.6 Sol
- SimpleBenchGPT-5.6 Sol by 15.964.848.9+15.9
- ARC-AGI-2GPT-5.6 Sol by 8.692.583.9+8.6
- ARC-AGI-3GPT-5.6 Sol by 77.80.8+7
- Humanity's Last ExamGPT-5.6 Sol by 6.649.542.9+6.6
- CritPtGPT-5.6 Sol by 2.332.330+2.3
- GPQA DiamondGPT-5.6 Sol by 1.694.192.5+1.6
- LiveBench · ReasoningGPT-5.6 Sol by 1.191.790.6+1.1
FactsGPT-5.6 Sol
- SimpleQA VerifiedGPT-5.6 Sol by 26.569.743.2+26.5
- AA-Omniscience · AccuracyGPT-5.6 Sol by 12.659.446.8+12.6
- AA-Omniscience · Non-hallucinationGPT-5.6 Terra by 4.37.812.1+4.3
Images and chartsGPT-5.6 Sol
- MMMU-ProGPT-5.6 Sol by 2.783.480.7+2.7
- LMArena · Visiontie12791271tie
- CharXiv (RQ)tie85.885.9tie
MathGPT-5.6 Sol
- FrontierMath Tier 4GPT-5.6 Sol by 14.78368.3+14.7
- FrontierMath Tiers 1-3 (v2)GPT-5.6 Sol by 3.189.186+3.1
- LiveBench · MathematicsGPT-5.6 Sol by 1.396.294.9+1.3
Following instructionsGPT-5.6 Sol
- LiveBench · Instruction FollowingGPT-5.6 Sol by 7.371.864.6+7.3
- IFBenchGPT-5.6 Sol by 1.572.771.2+1.5
Other results29 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- AA-OmniscienceGPT-5.6 Sol by 21.9220.1+21.9
- GDPval-AA v2 EloGPT-5.6 Sol by 182 rating points17101528+182 rating
- Terminal-Bench 4.0GPT-5.6 Sol by 15.837.321.5+15.8
- Terminal-bench 3.0GPT-5.6 Sol by 13.634.420.8+13.6
- OSWorld 2.0GPT-5.6 Sol by 12.462.650.2+12.4
- OSWorld 2.0 (partial)GPT-5.6 Sol by 12.462.650.2+12.4
- GDP.PDF (All pass rate)GPT-5.6 Sol by 114029+11
- AA Agentic IndexGPT-5.6 Sol by 6.850.543.7+6.8
- GDP (Surge AI)GPT-5.6 Sol by 630.724.7+6
- AA IntelligenceGPT-5.6 Sol by 4.94742.1+4.9
- ToolathlonGPT-5.6 Sol by 4.95853.1+4.9
- livebench_languageGPT-5.6 Sol by 4.887.782.9+4.8
- FrontierMath (overall)GPT-5.6 Sol by 4.18984.9+4.1
- Artificial AnalysisGPT-5.6 Sol by 45955+4
- HLE-VerifiedGPT-5.6 Sol by 3.454.551.1+3.4
- LVBenchGPT-5.6 Sol by 3.282.178.9+3.2
- DeepSWEGPT-5.6 Sol by 3.172.769.6+3.1
- DeepSWE 1.1GPT-5.6 Sol by 37370+3
- HealthBench ProfessionalGPT-5.6 Sol by 2.860.557.7+2.8
- HealthBench Professional length-adjustedGPT-5.6 Sol by 2.860.557.7+2.8
- Agents' Last ExamGPT-5.6 Sol by 2.352.750.4+2.3
- τ²-Bench Telecom (AA run)GPT-5.6 Terra by 1.285.186.3+1.2
- Artificial Analysis Coding Indextie77.476.7tie
- livebench_data_analysistie79.879.3tie
- HealthBench Consensus length-adjustedtie95.595.1tie
- HealthBench Hard length-adjustedtie33.132.7tie
- ARC-AGI-1tie96.596.5tie
- HealthBenchtie5757tie
- HealthBench length-adjustedtie5757tie
Questions people ask
Which is better, GPT-5.6 Sol or GPT-5.6 Terra?
GPT-5.6 Sol wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. GPT-5.6 Terra wins none, but costs 42% less.
Which is better for coding?
GPT-5.6 Sol. It wins 6 of the 7 coding tests both models report; GPT-5.6 Terra wins none, and 1 is a tie.
Which is cheaper?
GPT-5.6 Sol costs $4.00 per million input tokens and $20.00 per million output tokens; GPT-5.6 Terra costs $2.00 and $12.00. That makes GPT-5.6 Terra about 42% cheaper for the same work.
How do you compare the two?
We use the 61 benchmark tests both models have published scores on. The verdict counts the 32 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 29 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.