Claude 3.5 Sonnet vs Claude Sonnet 4.5
Wins 0 of 4 areas
—
Wins 4 of 4 areas
Coding · Reasoning · Images and charts · Long documents
Claude Sonnet 4.5 is the stronger all-rounder.
Scores updated · 29 tests both models report · How we compare
Where each one wins
Tests won in each of the four 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 software03Claude Sonnet 4.53 of 3 tests
- ReasoningHard problems that need careful thinking03Claude Sonnet 4.53 of 3 tests
- Images and chartsUnderstanding pictures, charts and video03Claude Sonnet 4.53 of 3 tests
- Long documentsFinding answers in very long texts01Claude Sonnet 4.51 of 1 test
Long documents 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.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Claude Sonnet 4.5 pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+28.2points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+27.4points ahead
- Recent programming contest problemsLiveCodeBench v6+26.8points ahead
Where Claude 3.5 Sonnet pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 4.5
- SWE-bench VerifiedClaude Sonnet 4.5 by 28.24977.2+28.2
- LiveCodeBench v6Claude Sonnet 4.5 by 26.837.264+26.8
- SciCodeClaude Sonnet 4.5 by 14.131.645.7+14.1
ReasoningClaude Sonnet 4.5
- GPQA DiamondClaude Sonnet 4.5 by 27.45683.4+27.4
- SimpleBenchClaude Sonnet 4.5 by 26.827.554.3+26.8
- Humanity's Last ExamClaude Sonnet 4.5 by 14.63.217.8+14.6
Images and chartsClaude Sonnet 4.5
Long documentsClaude Sonnet 4.5
- LongBench v2Claude Sonnet 4.5 by 20.84161.8+20.8
Other results19 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.
- AIME 2025Claude Sonnet 4.5 by 83.73.387+83.7
- HMMT Feb. 2025Claude Sonnet 4.5 by 65.81.767.5+65.8
- LiveCodeBenchClaude Sonnet 4.5 by 38.232.871+38.2
- Artificial Analysis Coding IndexClaude Sonnet 4.5 by 26.12652.1+26.1
- TAU-bench (airline)Claude Sonnet 4.5 by 244670+24
- TAU-bench (retail)Claude Sonnet 4.5 by 1769.286.2+17
- AA IntelligenceClaude Sonnet 4.5 by 13.57.220.7+13.5
- RealWorldQAClaude Sonnet 4.5 by 10.260.170.3+10.2
- MMLU-ProClaude Sonnet 4.5 by 9.977.687.5+9.9
- MMMU (val) (Pass@1)Claude Sonnet 4.5 by 9.568.377.8+9.5
- AI2DClaude 3.5 Sonnet by 7.794.787+7.7
- MMLU-ReduxClaude Sonnet 4.5 by 6.788.995.6+6.7
- HarmBenchClaude 3.5 Sonnet by 6.198.192+6.1
- frontiermath_tier_4_v1Claude Sonnet 4.5 by 4.204.2+4.2
- bbqClaude Sonnet 4.5 by 494.998.9+4
- AIR-Bench 2024Claude Sonnet 4.5 by 3.985.989.8+3.9
- anthropic_red_teamClaude 3.5 Sonnet by 1.699.898.2+1.6
- XSTesttie95.696.4tie
- simple_safety_teststie100100tie
Questions people ask
Which is better, Claude 3.5 Sonnet or Claude Sonnet 4.5?
Claude Sonnet 4.5 wins all four areas where both have results: coding, reasoning, images and charts and long documents. Claude 3.5 Sonnet wins none.
Which is better for coding?
Claude Sonnet 4.5. It wins 3 of the 3 coding tests both models report; Claude 3.5 Sonnet wins none.
Which is cheaper?
Claude 3.5 Sonnet costs $3.00 per million input tokens and $15.00 per million output tokens; Claude Sonnet 4.5 costs $3.00 and $15.00. For a million tokens read plus a million written, they cost about the same.
How do you compare the two?
We use the 29 benchmark tests both models have published scores on. The verdict counts the 10 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 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.