Claude Sonnet 4.6 vs Claude Sonnet 5
Wins 0 of 8 areas
—
Wins 6 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math
Claude Sonnet 5 is the stronger all-rounder.
Scores updated · 50 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 software07Claude Sonnet 57 of 7 tests
- AgentsCarrying out multi-step tasks on its own05Claude Sonnet 55 of 5 tests
- ReasoningHard problems that need careful thinking04Claude Sonnet 54 of 4 tests
- Images and chartsUnderstanding pictures, charts and video02Claude Sonnet 52 of 3 tests · 1 tie
- MathCompetition and research-level math02Claude Sonnet 52 of 2 tests
- FactsGetting facts right instead of making them up01Claude Sonnet 51 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts11Even1 each
- Following instructionsDoing exactly what it is asked00Even0 each · 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 33% 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 Claude Sonnet 5 pulls ahead
- Proof problems from the US Math OlympiadUSAMO 2026+24.5points ahead
- Fixes real GitHub issues, working as an agentLiveBench · Agentic Coding+16.8points ahead
- Reasoning about charts from research papersCharXiv (RQ)+16.7points ahead
Where Claude Sonnet 4.6 pulls ahead
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+3.4points ahead
Every test, side by side
All 50 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 5
- LiveBench · Agentic CodingClaude Sonnet 5 by 16.842.659.4+16.8
- Terminal-Bench 2.1Claude Sonnet 5 by 9.371.280.5+9.3
- SWE-bench VerifiedClaude Sonnet 5 by 5.679.685.2+5.6
- SWE-bench ProClaude Sonnet 5 by 5.158.163.2+5.1
- SciCodeClaude Sonnet 5 by 4.250.154.3+4.2
- LMArena · WebDevClaude Sonnet 5 by 19 rating points15221541+19 rating
- LiveBench · CodingClaude Sonnet 5 by 1.479.380.7+1.4
AgentsClaude Sonnet 5
- GDPValClaude Sonnet 5 by 11.636.748.3+11.6
- Terminal-Bench 4.0Claude Sonnet 5 by 11.1314.1+11.1
- BrowseCompClaude Sonnet 5 by 1074.784.7+10
- OSWorld-VerifiedClaude Sonnet 5 by 8.772.581.2+8.7
- τ-Bench V3 · BankingClaude Sonnet 5 by 2.934.437.3+2.9
ReasoningClaude Sonnet 5
- CritPtClaude Sonnet 5 by 13.83.116.9+13.8
- Humanity's Last ExamClaude Sonnet 5 by 7.733.641.3+7.7
- LiveBench · ReasoningClaude Sonnet 5 by 3.984.888.7+3.9
- GPQA DiamondClaude Sonnet 5 by 3.687.591.1+3.6
FactsClaude Sonnet 5
- AA-Omniscience · Non-hallucinationClaude Sonnet 5 by 951.660.6+9
- SimpleQA Verifiedtie32.833.7tie
- AA-Omniscience · Accuracytie40.940tie
Images and chartsClaude Sonnet 5
- CharXiv (RQ)Claude Sonnet 5 by 16.771.688.3+16.7
- MMMU-ProClaude Sonnet 5 by 473.377.3+4
- LMArena · Visiontie12831275tie
MathClaude Sonnet 5
- USAMO 2026Claude Sonnet 5 by 24.55579.5+24.5
- LiveBench · MathematicsClaude Sonnet 5 by 5.98792.9+5.9
Long documentsEven
- MRCR v2 8 needle 128k (average)Claude Sonnet 4.6 by 3.484.981.5+3.4
- AA-LCRClaude Sonnet 5 by 28082+2
Following instructionsEven
- LiveBench · Instruction Followingtie63.263.9tie
Other results23 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.
- DeepSWE 1.1Claude Sonnet 5 by 243054+24
- GDPval-AA v2Claude Sonnet 5 by 223 rating points13951618+223 rating
- Terminal-Bench 2.0Claude Sonnet 5 by 21.359.180.4+21.3
- HealthBench ProfessionalClaude Sonnet 5 by 13.644.257.8+13.6
- ToolathlonClaude Sonnet 5 by 13.34154.3+13.3
- AA Agentic IndexClaude Sonnet 5 by 11.233.144.3+11.2
- ChartMuseum (without tools)Claude Sonnet 5 by 10.859.370.1+10.8
- HLE (with tools)Claude Sonnet 5 by 10.646.857.4+10.6
- Artificial Analysis Coding IndexClaude Sonnet 5 by 8.56371.5+8.5
- AA IntelligenceClaude Sonnet 5 by 8.130.138.2+8.1
- OfficeQAClaude Sonnet 5 by 6.468.775.1+6.4
- livebench_data_analysisClaude Sonnet 4.6 by 6.37871.7+6.3
- OfficeQA ProClaude Sonnet 5 by 653.459.4+6
- ChartMuseum (with tools)Claude Sonnet 5 by 5.880.986.7+5.8
- AA-OmniscienceClaude Sonnet 5 by 4.312.216.4+4.3
- CharXiv Reasoning (With tools)Claude Sonnet 5 by 385.388.3+3
- GDP (Surge AI)Claude Sonnet 5 by 378.681.6+3
- Toolathlon Pass@∞Claude Sonnet 5 by 2.73840.7+2.7
- livebench_languageClaude Sonnet 4.6 by 1.176.175+1.1
- Legal Agent Benchmark, Full Public Settie88.9tie
- Legal Agent Benchmarktie5.45.8tie
- BenchCAD Vision2Code (with tools)tie0.30.4tie
- BenchCAD Vision2Code (without tools)tie0.30.3tie
Questions people ask
Which is better, Claude Sonnet 4.6 or Claude Sonnet 5?
Claude Sonnet 5 wins six of the eight areas we test: coding, agents, reasoning, facts, images and charts and math. Claude Sonnet 4.6 wins none. They are level on long documents and following instructions.
Which is better for coding?
Claude Sonnet 5. It wins 7 of the 7 coding tests both models report; Claude Sonnet 4.6 wins none.
Which is cheaper?
Claude Sonnet 4.6 costs $3.00 per million input tokens and $15.00 per million output tokens; Claude Sonnet 5 costs $2.00 and $10.00. That makes Claude Sonnet 5 about 33% cheaper for the same work.
How do you compare the two?
We use the 50 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 23 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.