Claude Opus 5.5 vs Claude Sonnet 5
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
Wins 0 of 8 areas
—
Claude Opus 5.5 is the stronger all-rounder.Claude Sonnet 5 is cheaper.
Scores updated · 51 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 software60Claude Opus 5.56 of 6 tests
- ReasoningHard problems that need careful thinking40Claude Opus 5.54 of 4 tests
- AgentsCarrying out multi-step tasks on its own30Claude Opus 5.53 of 3 tests
- MathCompetition and research-level math30Claude Opus 5.53 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20Claude Opus 5.52 of 2 tests
- FactsGetting facts right instead of making them up21Claude Opus 5.52 of 3 tests
- Long documentsFinding answers in very long texts10Claude Opus 5.51 of 1 test
- Following instructionsDoing exactly what it is asked10Claude Opus 5.51 of 1 test
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 50% 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 Opus 5.5 pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+65.7points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+45.5points ahead
- Short factual questions, answered correctlySimpleQA Verified+38.5points ahead
Where Claude Sonnet 5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+19.2points ahead
Every test, side by side
All 51 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 5.5
- LMArena · WebDevClaude Opus 5.5 by 275 rating points18141539+275 rating
- SWE-bench ProClaude Opus 5.5 by 26.789.963.2+26.7
- SWE-bench MultilingualClaude Opus 5.5 by 15.693.978.3+15.6
- SciCodeClaude Opus 5.5 by 12.666.954.3+12.6
- LiveBench · Agentic CodingClaude Opus 5.5 by 12.371.759.4+12.3
- LiveBench · CodingClaude Opus 5.5 by 8.689.380.7+8.6
AgentsClaude Opus 5.5
- Terminal-Bench 4.0Claude Opus 5.5 by 45.559.614.1+45.5
- GDPValClaude Opus 5.5 by 20.168.348.2+20.1
- BrowseCompClaude Opus 5.5 by 6.190.884.7+6.1
ReasoningClaude Opus 5.5
- SimpleBenchClaude Opus 5.5 by 27.888.460.6+27.8
- Humanity's Last ExamClaude Opus 5.5 by 20.161.441.3+20.1
- CritPtClaude Opus 5.5 by 14.831.716.9+14.8
- LiveBench · ReasoningClaude Opus 5.5 by 3.592.288.7+3.5
FactsClaude Opus 5.5
- SimpleQA VerifiedClaude Opus 5.5 by 38.572.233.7+38.5
- AA-Omniscience · AccuracyClaude Opus 5.5 by 26.166.240+26.1
- AA-Omniscience · Non-hallucinationClaude Sonnet 5 by 19.241.460.6+19.2
Images and chartsClaude Opus 5.5
- MMMU-ProClaude Opus 5.5 by 10.487.777.3+10.4
- LMArena · VisionClaude Opus 5.5 by 52 rating points13241272+52 rating
MathClaude Opus 5.5
- FrontierMath Tier 4Claude Opus 5.5 by 65.79529.3+65.7
- FrontierMath Tiers 1-3 (v2)Claude Opus 5.5 by 25.691.265.6+25.6
- LiveBench · MathematicsClaude Opus 5.5 by 4.297.192.9+4.2
Following instructionsClaude Opus 5.5
- LiveBench · Instruction FollowingClaude Opus 5.5 by 1.865.763.9+1.8
Other results28 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 4.0Claude Opus 5.5 by 5466.412.4+54
- GDPval-AA 2.1Claude Opus 5.5 by 397 rating points18461449+397 rating
- OSWorld 2.0 (partial)Claude Opus 5.5 by 39.281.842.6+39.2
- SWE-bench MultimodalClaude Opus 5.5 by 33.361.428.1+33.3
- AA-OmniscienceClaude Opus 5.5 by 29.946.416.4+29.9
- OSWorld 2.1Claude Opus 5.5 by 24.881.857+24.8
- OSWorld 2.1 partialClaude Opus 5.5 by 24.881.857+24.8
- OSWorld 2.1 partial scoreClaude Opus 5.5 by 24.881.857+24.8
- OSWorld 2.1 strict pass rateClaude Opus 5.5 by 23.148.725.6+23.1
- DeepSWE 1.1Claude Opus 5.5 by 20.274.254+20.2
- AA IntelligenceClaude Opus 5.5 by 19.457.638.2+19.4
- LVBenchClaude Opus 5.5 by 15.283.768.5+15.2
- HealthBench Professional (raw score)Claude Opus 5.5 by 14.777.162.4+14.7
- Program BenchClaude Opus 5.5 by 13.991.277.3+13.9
- Humanity's Last Exam (w/ tools)Claude Opus 5.5 by 12.867.754.9+12.8
- livebench_languageClaude Opus 5.5 by 11.386.375+11.3
- HLE (with tools)Claude Opus 5.5 by 10.367.757.4+10.3
- HealthBench (raw score)Claude Opus 5.5 by 8.968.159.2+8.9
- livebench_data_analysisClaude Opus 5.5 by 8.680.371.7+8.6
- OfficeQA ProClaude Opus 5.5 by 8.367.759.4+8.3
- HealthBench ProfessionalClaude Opus 5.5 by 7.865.657.8+7.8
- GMMLUClaude Opus 5.5 by 5.394.389+5.3
- Agent's Last Exam (Pass rate)Claude Opus 5.5 by 4.938.233.3+4.9
- MILUClaude Opus 5.5 by 3.893.189.3+3.8
- OfficeQAClaude Opus 5.5 by 3.878.975.1+3.8
- Toolathlon VerifiedClaude Opus 5.5 by 3.177.874.7+3.1
- Legal Agent BenchmarkClaude Opus 5.5 by 2.58.35.8+2.5
- HealthBenchClaude Opus 5.5 by 1.460.659.2+1.4
Questions people ask
Which is better, Claude Opus 5.5 or Claude Sonnet 5?
Claude Opus 5.5 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Claude Sonnet 5 wins none, but costs 50% less.
Which is better for coding?
Claude Opus 5.5. It wins 6 of the 6 coding tests both models report; Claude Sonnet 5 wins none.
Which is cheaper?
Claude Opus 5.5 costs $4.00 per million input tokens and $20.00 per million output tokens; Claude Sonnet 5 costs $2.00 and $10.00. That makes Claude Sonnet 5 about 50% cheaper for the same work.
How do you compare the two?
We use the 51 benchmark tests both models have published scores on. The verdict counts the 23 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 28 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.