Claude Sonnet 4.6 vs GPT-5.5
Wins 0 of 8 areas
—
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
GPT-5.5 is the stronger all-rounder.Claude Sonnet 4.6 is cheaper.
Scores updated · 58 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.
- AgentsCarrying out multi-step tasks on its own08GPT-5.58 of 8 tests
- CodingWriting and fixing software05GPT-5.55 of 7 tests · 2 ties
- ReasoningHard problems that need careful thinking05GPT-5.55 of 5 tests
- FactsGetting facts right instead of making them up13GPT-5.53 of 4 tests
- MathCompetition and research-level math02GPT-5.52 of 2 tests
- Following instructionsDoing exactly what it is asked02GPT-5.52 of 2 tests
- Images and chartsUnderstanding pictures, charts and video01GPT-5.51 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts01GPT-5.51 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.
Claude Sonnet 4.6 costs 49% 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-5.5 pulls ahead
- Proof problems from the US Math OlympiadUSAMO 2026+43.2points ahead
- Short factual questions, answered correctlySimpleQA Verified+30.2points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+26.7points ahead
Where Claude Sonnet 4.6 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+40.6points ahead
Every test, side by side
All 58 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.5
- Terminal-Bench 2.1GPT-5.5 by 13.171.284.3+13.1
- LiveBench · Agentic CodingGPT-5.5 by 11.442.654+11.4
- Terminal-Bench HardGPT-5.5 by 7.65360.6+7.6
- SciCodeGPT-5.5 by 5.750.155.8+5.7
- LiveBench · CodingGPT-5.5 by 2.979.382.2+2.9
- LMArena · WebDevtie15221513tie
- SWE-bench Protie58.158.6tie
AgentsGPT-5.5
- MCP AtlasGPT-5.5 by 1461.375.3+14
- Terminal-Bench 4.0GPT-5.5 by 11.6314.6+11.6
- AA ApexAgentsGPT-5.5 by 9.72837.7+9.7
- BrowseCompGPT-5.5 by 9.774.784.4+9.7
- OSWorld-VerifiedGPT-5.5 by 6.272.578.7+6.2
- AA IT-Bench SREGPT-5.5 by 639.845.8+6
- GDPValGPT-5.5 by 636.742.7+6
- τ-Bench V3 · BankingGPT-5.5 by 4.634.439+4.6
ReasoningGPT-5.5
- ARC-AGI-2GPT-5.5 by 26.758.385+26.7
- CritPtGPT-5.5 by 243.127.1+24
- Humanity's Last ExamGPT-5.5 by 12.233.645.8+12.2
- GPQA DiamondGPT-5.5 by 687.593.5+6
- LiveBench · ReasoningGPT-5.5 by 4.984.889.7+4.9
FactsGPT-5.5
- AA-Omniscience · Non-hallucinationClaude Sonnet 4.6 by 40.651.611+40.6
- SimpleQA VerifiedGPT-5.5 by 30.232.863+30.2
- AA-Omniscience · AccuracyGPT-5.5 by 17.140.958+17.1
- Vectara HHEM hallucination ratelower is betterGPT-5.5 by 1.310.69.3+1.3
Images and chartsGPT-5.5
- MMMU-ProGPT-5.5 by 6.673.379.9+6.6
- LMArena · Visiontie12831292tie
MathGPT-5.5
- USAMO 2026GPT-5.5 by 43.25598.2+43.2
- LiveBench · MathematicsGPT-5.5 by 8.98795.9+8.9
Following instructionsGPT-5.5
- IFBenchGPT-5.5 by 19.356.675.9+19.3
- LiveBench · Instruction FollowingGPT-5.5 by 7.563.270.7+7.5
Other results27 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.
- GDP (Surge AI)Claude Sonnet 4.6 by 53.778.624.9+53.7
- DeepSWE 1.1GPT-5.5 by 373067+37
- frontiermath_tier_4_v1GPT-5.5 by 27.18.335.4+27.1
- Terminal-Bench 2.0GPT-5.5 by 23.659.182.7+23.6
- τ²-Bench Telecom (AA run)GPT-5.5 by 18.275.793.9+18.2
- ToolathlonGPT-5.5 by 14.64155.6+14.6
- Artificial Analysis Coding IndexGPT-5.5 by 11.96374.9+11.9
- GDPval-AA v2GPT-5.5 by 114 rating points13951509+114 rating
- livebench_languageGPT-5.5 by 11.376.187.4+11.3
- GDPval-AA (Elo)GPT-5.5 by 93 rating points16761769+93 rating
- ARC-AGI-1GPT-5.5 by 98695+9
- AA IntelligenceGPT-5.5 by 8.330.138.4+8.3
- AA-OmniscienceGPT-5.5 by 8.312.220.5+8.3
- BrowseComp (single-agent)GPT-5.5 by 8.276.284.4+8.2
- HealthBench ProfessionalGPT-5.5 by 7.644.251.8+7.6
- HLE (with tools)GPT-5.5 by 5.446.852.2+5.4
- LiveBenchGPT-5.5 by 5.275.580.7+5.2
- AA Agentic IndexGPT-5.5 by 4.233.137.3+4.2
- livebench_data_analysisGPT-5.5 by 3.67881.6+3.6
- Finance AgentClaude Sonnet 4.6 by 3.363.360+3.3
- Legal Agent BenchmarkClaude Sonnet 4.6 by 3.35.42.1+3.3
- Legal Agent Benchmark (Harvey's Held-Out Set)Claude Sonnet 4.6 by 3.35.42.1+3.3
- vectara_avg_summary_lengthGPT-5.5 by 14.9 rating points114.7129.6+14.9 rating
- vectara_factual_consistencyGPT-5.5 by 1.389.490.7+1.3
- Finance Agent v2tie5151.8tie
- OfficeQA Protie53.454.1tie
- vectara_answer_ratetie99.9100tie
Questions people ask
Which is better, Claude Sonnet 4.6 or GPT-5.5?
GPT-5.5 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Claude Sonnet 4.6 wins none, but costs 49% less.
Which is better for coding?
GPT-5.5. It wins 5 of the 7 coding tests both models report; Claude Sonnet 4.6 wins none, and 2 are ties.
Which is cheaper?
Claude Sonnet 4.6 costs $3.00 per million input tokens and $15.00 per million output tokens; GPT-5.5 costs $5.00 and $30.00. That makes Claude Sonnet 4.6 about 49% cheaper for the same work.
How do you compare the two?
We use the 58 benchmark tests both models have published scores on. The verdict counts the 31 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 27 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.