Claude Opus 4 vs o4-mini
Wins 4 of 6 areas
Coding · Reasoning · Facts · Long documents
Wins 1 of 6 areas
Images and charts
Claude Opus 4 is the stronger all-rounder.o4-mini is cheaper and better at images and charts.
Scores updated · 34 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 software21Claude Opus 42 of 3 tests
- ReasoningHard problems that need careful thinking21Claude Opus 42 of 3 tests
- FactsGetting facts right instead of making them up10Claude Opus 41 of 1 test
- Long documentsFinding answers in very long texts10Claude Opus 41 of 1 test
- Images and chartsUnderstanding pictures, charts and video01o4-mini1 of 2 tests · 1 tie
- 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.
o4-mini costs 94% 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 Opus 4 pulls ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+15.9points ahead
- Keeps track of context across a multi-turn chatMulti-Challenge+15.6points ahead
- Reasons across sets of long documentsAA-LCR+8.3points ahead
Every test, side by side
All 34 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4
- Terminal-Bench HardClaude Opus 4 by 15.931.115.2+15.9
- SciCodeo4-mini by 6.739.846.5+6.7
- SWE-bench VerifiedClaude Opus 4 by 4.472.568.1+4.4
ReasoningClaude Opus 4
- Humanity's Last Examo4-mini by 4.212.316.5+4.2
- ARC-AGI-2Claude Opus 4 by 2.58.66.1+2.5
- GPQA DiamondClaude Opus 4 by 1.279.678.4+1.2
FactsClaude Opus 4
- Vectara HHEM hallucination ratelower is betterClaude Opus 4 by 6.61218.6+6.6
Images and chartso4-mini
- MMMUo4-mini by 7.973.781.6+7.9
- LMArena · Visiontie11921194tie
Following instructionsEven
- Multi-ChallengeClaude Opus 4 by 15.658.643+15.6
- IFBencho4-mini by 1553.768.7+15
Other results22 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.
- ARC-AGI-1o4-mini by 2335.758.7+23
- τ²-Bench Telecom (AA run)Claude Opus 4 by 17.873.455.6+17.8
- AIME 2025o4-mini by 17.275.592.7+17.2
- livecodebench_hardo4-mini by 15.833.148.9+15.8
- TAU-bench (airline)Claude Opus 4 by 10.459.649.2+10.4
- TAU-bench (retail)Claude Opus 4 by 9.681.471.8+9.6
- Artificial Analysis Coding IndexClaude Opus 4 by 8.43425.6+8.4
- vectara_answer_rateo4-mini by 8.29199.2+8.2
- AIR-Bench 2024Claude Opus 4 by 7.285.778.5+7.2
- LiveCodeBencho4-mini by 770.477.4+7
- vectara_factual_consistencyClaude Opus 4 by 6.68881.4+6.6
- livecodebench_mediumo4-mini by 5.880.486.2+5.8
- HarmBencho4-mini by 5.391.797+5.3
- AA IntelligenceClaude Opus 4 by 3.920.616.7+3.9
- bbqClaude Opus 4 by 3.197.194+3.1
- frontiermath_tier_4_v1o4-mini by 2.14.26.3+2.1
- Aider-PolyglotClaude Opus 4 by 1.870.768.9+1.8
- anthropic_red_teamtie98.998.2tie
- vectara_avg_summary_lengthtie123.2127.7tie
- XSTesttie9797.4tie
- livecodebench_easytie99.198.8tie
- simple_safety_teststie100100tie
Questions people ask
Which is better, Claude Opus 4 or o4-mini?
Claude Opus 4 wins four of the six areas where both have results: coding, reasoning, facts and long documents. o4-mini wins images and charts, and costs 94% less. They are level on following instructions.
Which is better for coding?
Claude Opus 4. It wins 2 of the 3 coding tests both models report; o4-mini wins 1.
Which is cheaper?
Claude Opus 4 costs $15.00 per million input tokens and $75.00 per million output tokens; o4-mini costs $1.10 and $4.40. That makes o4-mini about 94% cheaper for the same work.
How do you compare the two?
We use the 34 benchmark tests both models have published scores on. The verdict counts the 12 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 22 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.