GPT-4o vs o4-mini
Wins 1 of 7 areas
Facts
Wins 6 of 7 areas
Coding · Agents · Reasoning · Images and charts · Long documents · Following instructions
o4-mini is the stronger all-rounder.GPT-4o is better at facts.
Scores updated · 43 tests both models report · How we compare
Where each one wins
Tests won in each of the seven 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 video04o4-mini4 of 4 tests
- ReasoningHard problems that need careful thinking03o4-mini3 of 4 tests · 1 tie
- CodingWriting and fixing software03o4-mini3 of 3 tests
- Following instructionsDoing exactly what it is asked02o4-mini2 of 2 tests
- AgentsCarrying out multi-step tasks on its own01o4-mini1 of 1 test
- Long documentsFinding answers in very long texts01o4-mini1 of 1 test
- FactsGetting facts right instead of making them up31GPT-4o3 of 4 tests
Agents 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 73% 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 o4-mini pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+34.9points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+32.7points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+25.8points ahead
Where GPT-4o pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+42.6points ahead
- Short factual questions, answered correctlySimpleQA Verified+7.2points ahead
Every test, side by side
All 43 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Codingo4-mini
- SWE-bench Verifiedo4-mini by 34.933.268.1+34.9
- SciCodeo4-mini by 13.233.346.5+13.2
- Terminal-Bench Hardo4-mini by 6.98.315.2+6.9
Reasoningo4-mini
- GPQA Diamondo4-mini by 25.852.678.4+25.8
- Humanity's Last Examo4-mini by 14.71.816.5+14.7
- ARC-AGI-2o4-mini by 6.106.1+6.1
- CritPttie00.6tie
FactsGPT-4o
- AA-Omniscience · Non-hallucinationGPT-4o by 42.662.119.5+42.6
- Vectara HHEM hallucination ratelower is betterGPT-4o by 99.618.6+9
- SimpleQA VerifiedGPT-4o by 7.22618.8+7.2
- AA-Omniscience · Accuracyo4-mini by 1.123.724.8+1.1
Images and chartso4-mini
- MathVistao4-mini by 22.961.484.3+22.9
- CharXiv (RQ)o4-mini by 13.258.872+13.2
- MMMU-Proo4-mini by 12.956.369.2+12.9
- MMMUo4-mini by 9.472.281.6+9.4
Following instructionso4-mini
- IFBencho4-mini by 32.73668.7+32.7
- Multi-Challengeo4-mini by 2.740.343+2.7
Other results24 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 2025o4-mini by 8111.792.7+81
- ARC-AGI-1o4-mini by 54.24.558.7+54.2
- livecodebench_mediumo4-mini by 54.132.186.2+54.1
- livecodebench_hardo4-mini by 44.44.548.9+44.4
- LiveCodeBencho4-mini by 39.138.377.4+39.1
- Aider-Polygloto4-mini by 38.230.768.9+38.2
- AA Agentic Indexo4-mini by 27.78.436.1+27.7
- τ²-Bench Telecom (AA run)o4-mini by 26.728.955.6+26.7
- AA-OmniscienceGPT-4o by 25.2-10.5-35.7+25.2
- livecodebench_easyo4-mini by 16.382.598.8+16.3
- AIR-Bench 2024o4-mini by 16.162.478.5+16.1
- HarmBencho4-mini by 14.182.997+14.1
- TAU-bench (retail)o4-mini by 11.560.371.8+11.5
- AA Intelligenceo4-mini by 9.47.316.7+9.4
- vectara_factual_consistencyGPT-4o by 990.481.4+9
- MMLU-ProXo4-mini by 8.261.169.3+8.2
- TAU-bench (airline)o4-mini by 6.442.849.2+6.4
- vectara_answer_rateo4-mini by 5.493.899.2+5.4
- vectara_avg_summary_lengtho4-mini by 41.1 rating points86.6127.7+41.1 rating
- simple_safety_testso4-mini by 1.598.5100+1.5
- Artificial Analysis Coding Indexo4-mini by 1.424.225.6+1.4
- bbqGPT-4o by 1.195.194+1.1
- anthropic_red_teamtie99.198.2tie
- XSTesttie97.397.4tie
Questions people ask
Which is better, GPT-4o or o4-mini?
o4-mini wins six of the seven areas where both have results: coding, agents, reasoning, images and charts, long documents and following instructions. GPT-4o wins facts.
Which is better for coding?
o4-mini. It wins 3 of the 3 coding tests both models report; GPT-4o wins none.
Which is cheaper?
GPT-4o costs $5.00 per million input tokens and $15.00 per million output tokens; o4-mini costs $1.10 and $4.40. That makes o4-mini about 73% cheaper for the same work.
How do you compare the two?
We use the 43 benchmark tests both models have published scores on. The verdict counts the 19 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 24 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.