Claude Opus 4 vs Claude Sonnet 4
Wins 1 of 6 areas
Reasoning
Wins 3 of 6 areas
Coding · Facts · Long documents
Claude Sonnet 4 wins more areas, narrowly.Claude Opus 4 is better at reasoning.
Scores updated · 59 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. 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 software01Claude Sonnet 41 of 4 tests · 3 ties
- FactsGetting facts right instead of making them up01Claude Sonnet 41 of 1 test
- Long documentsFinding answers in very long texts01Claude Sonnet 41 of 1 test
- ReasoningHard problems that need careful thinking40Claude Opus 44 of 4 tests
- Images and chartsUnderstanding pictures, charts and video00Even0 each · 2 ties
- Following instructionsDoing exactly what it is asked11Even1 each
Facts and long documents 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 4 costs 80% 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 4 pulls ahead
- Recent programming contest problemsLiveCodeBench v6+1.1points ahead
- Reasons across sets of long documentsAA-LCR+1points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+1points ahead
Where Claude Opus 4 pulls ahead
- Common-sense trick questionsSimpleBench+13.3points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+2.7points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+1.9points ahead
Every test, side by side
All 59 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 4
- LiveCodeBench v6Claude Sonnet 4 by 1.147.448.5+1.1
- SciCodetie39.840tie
- SWE-bench Verifiedtie72.572.7tie
- Terminal-Bench Hardtie31.131.1tie
ReasoningClaude Opus 4
- SimpleBenchClaude Opus 4 by 13.358.845.5+13.3
- ARC-AGI-2Claude Opus 4 by 2.78.65.9+2.7
- GPQA DiamondClaude Opus 4 by 1.979.677.7+1.9
- Humanity's Last ExamClaude Opus 4 by 1.612.310.7+1.6
FactsClaude Sonnet 4
- Vectara HHEM hallucination ratelower is betterClaude Sonnet 4 by 1.71210.3+1.7
Following instructionsEven
- Multi-ChallengeClaude Opus 4 by 1.558.657.1+1.5
- IFBenchClaude Sonnet 4 by 153.754.7+1
Other results45 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.
- ZebraLogicClaude Sonnet 4 by 14.459.373.7+14.4
- τ²-BenchClaude Opus 4 by 11.85745.2+11.8
- τ²-Bench Telecom (AA run)Claude Opus 4 by 8.873.464.6+8.8
- Terminal-BenchClaude Opus 4 by 7.743.235.5+7.7
- vectara_answer_rateClaude Sonnet 4 by 7.69198.6+7.6
- SimpleQAClaude Opus 4 by 6.922.815.9+6.9
- τ²-Bench (Retail)Claude Opus 4 by 6.881.875+6.8
- HarmBenchClaude Sonnet 4 by 6.391.798+6.3
- AIME 2025Claude Opus 4 by 575.570.5+5
- AIME 2024Claude Opus 4 by 4.848.243.4+4.8
- Tau2 airlineClaude Opus 4 by 4.56055.5+4.5
- livecodebench_mediumClaude Opus 4 by 4.480.476+4.4
- ARC-AGI-1Claude Sonnet 4 by 4.335.740+4.3
- OJBenchClaude Opus 4 by 4.319.615.3+4.3
- frontiermath_tier_4_v1Claude Opus 4 by 4.24.20+4.2
- AutoLogiClaude Sonnet 4 by 3.786.189.8+3.7
- Artificial Analysis Coding IndexClaude Sonnet 4 by 3.63437.6+3.6
- PolyMath-enClaude Sonnet 4 by 349.852.8+3
- MMLU-ProClaude Opus 4 by 2.986.683.7+2.9
- CNMO 2024Claude Sonnet 4 by 2.857.660.4+2.8
- AIR-Bench 2024Claude Sonnet 4 by 2.685.788.3+2.6
- MMMLUClaude Opus 4 by 2.388.886.5+2.3
- vectara_avg_summary_lengthClaude Sonnet 4 by 22.6 rating points123.2145.8+22.6 rating
- LiveCodeBenchClaude Opus 4 by 1.970.468.5+1.9
- AA IntelligenceClaude Opus 4 by 1.720.618.9+1.7
- vectara_factual_consistencyClaude Sonnet 4 by 1.78889.7+1.7
- MMLUClaude Opus 4 by 1.492.991.5+1.4
- MultiPL-EClaude Opus 4 by 189.688.6+1
- TAU-bench (retail)tie81.480.5tie
- AIMEtie33.933.1tie
- SuperGPQAtie56.555.7tie
- AceBenchtie75.676.2tie
- MMLU-Reduxtie94.293.6tie
- bbqtie97.196.6tie
- livecodebench_hardtie33.132.6tie
- MATH-500 (EM)tie94.494tie
- TAU-bench (airline)tie59.660tie
- XSTesttie9796.6tie
- anthropic_red_teamtie98.999.2tie
- livecodebench_easytie99.198.8tie
- IFEvaltie87.487.6tie
- LiveBenchtie74.674.8tie
- Aider-Polyglottie70.770.7tie
- HMMT 2025tie15.915.9tie
- simple_safety_teststie100100tie
Questions people ask
Which is better, Claude Opus 4 or Claude Sonnet 4?
Claude Sonnet 4 wins three of the six areas where both have results: coding, facts and long documents. Claude Opus 4 wins reasoning. They are level on images and charts and following instructions.
Which is better for coding?
Claude Sonnet 4. It wins 1 of the 4 coding tests both models report; Claude Opus 4 wins none, and 3 are ties.
Which is cheaper?
Claude Opus 4 costs $15.00 per million input tokens and $75.00 per million output tokens; Claude Sonnet 4 costs $3.00 and $15.00. That makes Claude Sonnet 4 about 80% cheaper for the same work.
How do you compare the two?
We use the 59 benchmark tests both models have published scores on. The verdict counts the 14 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 45 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.