Claude Opus 4 vs o3
Wins 0 of 5 areas
—
Wins 4 of 5 areas
Coding · Images and charts · Long documents · Following instructions
o3 is the stronger all-rounder.
Scores updated · 44 tests both models report · How we compare
Where each one wins
Tests won in each of the five 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.
- Images and chartsUnderstanding pictures, charts and video02o32 of 2 tests
- Long documentsFinding answers in very long texts02o32 of 2 tests
- Following instructionsDoing exactly what it is asked02o32 of 2 tests
- CodingWriting and fixing software12o32 of 3 tests
- ReasoningHard problems that need careful thinking22Even2 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.
o3 costs 89% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where o3 pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+17.7points ahead
- College exam questions with charts, maps and diagramsMMMU+9.2points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+7.8points ahead
Where Claude Opus 4 pulls ahead
- Common-sense trick questionsSimpleBench+5.7points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+3.4points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+2.1points ahead
Every test, side by side
All 44 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Codingo3
- Terminal-Bench Hardo3 by 631.137.1+6
- SWE-bench VerifiedClaude Opus 4 by 3.472.569.1+3.4
- SciCodeo3 by 1.239.841+1.2
ReasoningEven
- Humanity's Last Examo3 by 7.812.320.1+7.8
- SimpleBenchClaude Opus 4 by 5.758.853.1+5.7
- GPQA Diamondo3 by 3.179.682.7+3.1
- ARC-AGI-2Claude Opus 4 by 2.18.66.5+2.1
Images and chartso3
- MMMUo3 by 9.273.782.9+9.2
- LMArena · Visiono3 by 22 rating points11921214+22 rating
Long documentso3
- AA-LCRo3 by 5.469.374.7+5.4
- LongBench v2o3 by 3.255.658.8+3.2
Following instructionso3
- IFBencho3 by 17.753.771.4+17.7
- Multi-Challengeo3 by 1.858.660.4+1.8
Other results31 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 2024o3 by 43.448.291.6+43.4
- ZebraLogico3 by 36.559.395.8+36.5
- livecodebench_hardo3 by 32.933.166+32.9
- SimpleQAo3 by 26.622.849.4+26.6
- ARC-AGI-1o3 by 25.135.760.8+25.1
- LiveCodeBench (24/8~25/5)o3 by 19.256.675.8+19.2
- HMMT Feb. 2025o3 by 17.56077.5+17.5
- LiveCodeBencho3 by 14.370.484.7+14.3
- AIME 2025o3 by 10.975.586.4+10.9
- Aider-Polygloto3 by 10.670.781.3+10.6
- livecodebench_mediumo3 by 9.480.489.8+9.4
- OpenAI-MRCR (128k)o3 by 7.648.956.5+7.6
- TAU-bench (airline)Claude Opus 4 by 7.659.652+7.6
- TAU-bench (retail)Claude Opus 4 by 7.581.473.9+7.5
- τ²-Bench Telecom (AA run)o3 by 7.373.480.7+7.3
- HarmBencho3 by 6.791.798.4+6.7
- metr_re_benchClaude Opus 4 by 52015+5
- Tau2 airlineo3 by 4.86064.8+4.8
- Artificial Analysis Coding Indexo3 by 4.43438.4+4.4
- MATH-500 (EM)o3 by 3.794.498.1+3.7
- frontiermath_tier_4_v1Claude Opus 4 by 2.14.22.1+2.1
- MMLU-ProClaude Opus 4 by 1.686.685+1.6
- τ²-Bench (Retail)Claude Opus 4 by 1.681.880.2+1.6
- AIR-Bench 2024Claude Opus 4 by 1.285.784.5+1.2
- FullStackBenchClaude Opus 4 by 170.369.3+1
- simple_safety_testsClaude Opus 4 by 110099+1
- bbqtie97.197.9tie
- anthropic_red_teamtie98.998.3tie
- AA Intelligencetie20.620.2tie
- XSTesttie9797.3tie
- livecodebench_easytie99.199.1tie
Questions people ask
Which is better, Claude Opus 4 or o3?
o3 wins four of the five areas where both have results: coding, images and charts, long documents and following instructions. Claude Opus 4 wins none. They are level on reasoning.
Which is better for coding?
o3. It wins 2 of the 3 coding tests both models report; Claude Opus 4 wins 1.
Which is cheaper?
Claude Opus 4 costs $15.00 per million input tokens and $75.00 per million output tokens; o3 costs $2.00 and $8.00. That makes o3 about 89% cheaper for the same work.
How do you compare the two?
We use the 44 benchmark tests both models have published scores on. The verdict counts the 13 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 31 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.