Claude Sonnet 5 vs GPT-5.5
Wins 1 of 8 areas
Agents
Wins 6 of 8 areas
Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
GPT-5.5 is the stronger all-rounder.Claude Sonnet 5 is cheaper and better at agents.
Scores updated · 45 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 thinking05GPT-5.55 of 5 tests
- MathCompetition and research-level math04GPT-5.54 of 4 tests
- Images and chartsUnderstanding pictures, charts and video02GPT-5.52 of 2 tests
- FactsGetting facts right instead of making them up12GPT-5.52 of 3 tests
- Long documentsFinding answers in very long texts01GPT-5.51 of 1 test
- Following instructionsDoing exactly what it is asked01GPT-5.51 of 1 test
- AgentsCarrying out multi-step tasks on its own21Claude Sonnet 52 of 5 tests · 2 ties
- CodingWriting and fixing software33Even3 each
Long documents and following instructions rest on a single test each.
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 5 costs 66% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5.5 pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+43.2points ahead
- Short factual questions, answered correctlySimpleQA Verified+29.3points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+19.7points ahead
Where Claude Sonnet 5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+49.6points ahead
- Real work tasks from 44 professionsGDPVal+5.6points ahead
- Fixes real GitHub issues, working as an agentLiveBench · Agentic Coding+5.4points ahead
Every test, side by side
All 45 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- LiveBench · Agentic CodingClaude Sonnet 5 by 5.459.454+5.4
- SWE-bench ProClaude Sonnet 5 by 4.663.258.6+4.6
- Terminal-Bench 2.1GPT-5.5 by 3.880.584.3+3.8
- LMArena · WebDevClaude Sonnet 5 by 28 rating points15411513+28 rating
- LiveBench · CodingGPT-5.5 by 1.580.782.2+1.5
- SciCodeGPT-5.5 by 1.554.355.8+1.5
AgentsClaude Sonnet 5
- GDPValClaude Sonnet 5 by 5.648.342.7+5.6
- OSWorld-VerifiedClaude Sonnet 5 by 2.581.278.7+2.5
- τ-Bench V3 · BankingGPT-5.5 by 1.737.339+1.7
- Terminal-Bench 4.0tie14.114.6tie
- BrowseComptie84.784.4tie
ReasoningGPT-5.5
- CritPtGPT-5.5 by 10.216.927.1+10.2
- SimpleBenchGPT-5.5 by 8.460.669+8.4
- Humanity's Last ExamGPT-5.5 by 4.541.345.8+4.5
- GPQA DiamondGPT-5.5 by 2.491.193.5+2.4
- LiveBench · ReasoningGPT-5.5 by 188.789.7+1
FactsGPT-5.5
- AA-Omniscience · Non-hallucinationClaude Sonnet 5 by 49.660.611+49.6
- SimpleQA VerifiedGPT-5.5 by 29.333.763+29.3
- AA-Omniscience · AccuracyGPT-5.5 by 17.94058+17.9
Images and chartsGPT-5.5
- MMMU-ProGPT-5.5 by 2.677.379.9+2.6
- LMArena · VisionGPT-5.5 by 17 rating points12751292+17 rating
MathGPT-5.5
- FrontierMath Tier 4GPT-5.5 by 43.229.372.5+43.2
- FrontierMath Tiers 1-3 (v2)GPT-5.5 by 19.765.685.3+19.7
- USAMO 2026GPT-5.5 by 18.779.598.2+18.7
- LiveBench · MathematicsGPT-5.5 by 392.995.9+3
Following instructionsGPT-5.5
- LiveBench · Instruction FollowingGPT-5.5 by 6.863.970.7+6.8
Other results18 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 5 by 56.781.624.9+56.7
- DeepSWE 1.1GPT-5.5 by 135467+13
- livebench_languageGPT-5.5 by 12.47587.4+12.4
- GDPval-AA v2Claude Sonnet 5 by 109 rating points16181509+109 rating
- livebench_data_analysisGPT-5.5 by 9.971.781.6+9.9
- GDPval-AA v2 EloClaude Sonnet 5 by 93 rating points15841491+93 rating
- AA Agentic IndexClaude Sonnet 5 by 744.337.3+7
- Program BenchClaude Sonnet 5 by 6.577.370.8+6.5
- HealthBench ProfessionalClaude Sonnet 5 by 657.851.8+6
- OfficeQA ProClaude Sonnet 5 by 5.359.454.1+5.3
- HLE (with tools)Claude Sonnet 5 by 5.257.452.2+5.2
- AA-OmniscienceGPT-5.5 by 416.420.5+4
- Legal Agent BenchmarkClaude Sonnet 5 by 3.75.82.1+3.7
- Artificial Analysis Coding IndexGPT-5.5 by 3.471.574.9+3.4
- Terminal-Bench 2.0GPT-5.5 by 2.380.482.7+2.3
- ToolathlonGPT-5.5 by 1.354.355.6+1.3
- Toolathlon VerifiedClaude Sonnet 5 by 1.274.773.5+1.2
- AA Intelligencetie38.238.4tie
Questions people ask
Which is better, Claude Sonnet 5 or GPT-5.5?
GPT-5.5 wins six of the eight areas we test: reasoning, facts, images and charts, math, long documents and following instructions. Claude Sonnet 5 wins agents, and costs 66% less. They are level on coding.
Which is better for coding?
Neither. They win 3 coding tests each of the 6 both models report.
Which is cheaper?
Claude Sonnet 5 costs $2.00 per million input tokens and $10.00 per million output tokens; GPT-5.5 costs $5.00 and $30.00. That makes Claude Sonnet 5 about 66% cheaper for the same work.
How do you compare the two?
We use the 45 benchmark tests both models have published scores on. The verdict counts the 27 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 18 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.