Claude Sonnet 5 vs GPT-5.2
Wins 5 of 8 areas
Coding · Agents · Reasoning · Images and charts · Following instructions
Wins 3 of 8 areas
Facts · Math · Long documents
Claude Sonnet 5 is the stronger all-rounder.GPT-5.2 is better at math.
Scores updated · 36 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.
- CodingWriting and fixing software70Claude Sonnet 57 of 7 tests
- ReasoningHard problems that need careful thinking40Claude Sonnet 54 of 5 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video30Claude Sonnet 53 of 3 tests
- AgentsCarrying out multi-step tasks on its own20Claude Sonnet 52 of 3 tests · 1 tie
- Following instructionsDoing exactly what it is asked10Claude Sonnet 51 of 1 test
- MathCompetition and research-level math02GPT-5.22 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up12GPT-5.22 of 3 tests
- Long documentsFinding answers in very long texts01GPT-5.21 of 2 tests · 1 tie
Following instructions 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 5 costs 24% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Claude Sonnet 5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+41.8points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+33.3points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+26.2points ahead
Where GPT-5.2 pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+4.2points ahead
- Short factual questions, answered correctlySimpleQA Verified+3.4points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+2.4points ahead
Every test, side by side
All 36 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 5
- SWE-bench ProClaude Sonnet 5 by 33.363.229.9+33.3
- LMArena · WebDevClaude Sonnet 5 by 125 rating points15411416+125 rating
- SWE-bench MultilingualClaude Sonnet 5 by 11.678.366.7+11.6
- LiveBench · Agentic CodingClaude Sonnet 5 by 9.159.450.3+9.1
- SWE-bench VerifiedClaude Sonnet 5 by 5.285.280+5.2
- LiveBench · CodingClaude Sonnet 5 by 4.680.776.1+4.6
- SciCodeClaude Sonnet 5 by 2.254.352.1+2.2
AgentsClaude Sonnet 5
- τ-Bench V3 · BankingClaude Sonnet 5 by 26.237.311.1+26.2
- BrowseCompClaude Sonnet 5 by 18.984.765.8+18.9
- GDPValtie48.348.3tie
ReasoningClaude Sonnet 5
- SimpleBenchClaude Sonnet 5 by 14.860.645.8+14.8
- LiveBench · ReasoningClaude Sonnet 5 by 5.588.783.2+5.5
- CritPtClaude Sonnet 5 by 5.316.911.6+5.3
- Humanity's Last ExamClaude Sonnet 5 by 3.641.337.7+3.6
- GPQA Diamondtie91.190.3tie
FactsGPT-5.2
- AA-Omniscience · Non-hallucinationClaude Sonnet 5 by 41.860.618.8+41.8
- AA-Omniscience · AccuracyGPT-5.2 by 4.24044.3+4.2
- SimpleQA VerifiedGPT-5.2 by 3.433.737.1+3.4
Images and chartsClaude Sonnet 5
- CharXiv (RQ)Claude Sonnet 5 by 6.288.382.1+6.2
- LMArena · VisionClaude Sonnet 5 by 29 rating points12751246+29 rating
- MMMU-ProClaude Sonnet 5 by 2.777.374.6+2.7
MathGPT-5.2
- FrontierMath Tier 4GPT-5.2 by 2.429.331.7+2.4
- FrontierMath Tiers 1-3 (v2)GPT-5.2 by 1.865.667.4+1.8
- LiveBench · Mathematicstie92.993.2tie
Long documentsGPT-5.2
- MRCR v2 8 needle 128k (average)GPT-5.2 by 2.381.583.8+2.3
- AA-LCRtie8282.7tie
Following instructionsClaude Sonnet 5
- LiveBench · Instruction FollowingClaude Sonnet 5 by 2.163.961.8+2.1
Other results9 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.
- Terminal-Bench 2.0Claude Sonnet 5 by 26.480.454+26.4
- Artificial Analysis Coding IndexClaude Sonnet 5 by 22.871.548.7+22.8
- AA-OmniscienceClaude Sonnet 5 by 17.416.4-0.9+17.4
- AA Agentic IndexGPT-5.2 by 15.944.360.2+15.9
- HLE (with tools)Claude Sonnet 5 by 11.957.445.5+11.9
- ToolathlonClaude Sonnet 5 by 854.346.3+8
- AA IntelligenceClaude Sonnet 5 by 7.838.230.4+7.8
- livebench_data_analysisGPT-5.2 by 6.571.778.2+6.5
- livebench_languageGPT-5.2 by 4.87579.8+4.8
Questions people ask
Which is better, Claude Sonnet 5 or GPT-5.2?
Claude Sonnet 5 wins five of the eight areas we test: coding, agents, reasoning, images and charts and following instructions. GPT-5.2 wins facts, math and long documents.
Which is better for coding?
Claude Sonnet 5. It wins 7 of the 7 coding tests both models report; GPT-5.2 wins none.
Which is cheaper?
Claude Sonnet 5 costs $2.00 per million input tokens and $10.00 per million output tokens; GPT-5.2 costs $1.75 and $14.00. That makes Claude Sonnet 5 about 24% cheaper for the same work.
How do you compare the two?
We use the 36 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 9 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.