GPT-4.1 mini vs o3
Wins 0 of 8 areas
—
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
o3 is the stronger all-rounder.GPT-4.1 mini is cheaper.
Scores updated · 36 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. 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 video05o35 of 5 tests
- ReasoningHard problems that need careful thinking04o34 of 4 tests
- FactsGetting facts right instead of making them up03o33 of 3 tests
- CodingWriting and fixing software02o32 of 3 tests · 1 tie
- Following instructionsDoing exactly what it is asked02o32 of 2 tests
- AgentsCarrying out multi-step tasks on its own01o31 of 1 test
- MathCompetition and research-level math01o31 of 1 test
- Long documentsFinding answers in very long texts01o31 of 1 test
Agents, math 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.
GPT-4.1 mini costs 80% 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 o3 pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+45.5points ahead
- Short factual questions, answered correctlySimpleQA Verified+36.7points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+33.1points ahead
Where GPT-4.1 mini pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 36 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Codingo3
- SWE-bench Verifiedo3 by 45.523.669.1+45.5
- Terminal-Bench Hardo3 by 29.57.637.1+29.5
- SciCodetie40.441tie
Reasoningo3
- GPQA Diamondo3 by 16.366.482.7+16.3
- Humanity's Last Examo3 by 15.1520.1+15.1
- ARC-AGI-2o3 by 6.506.5+6.5
- CritPto3 by 1.101.1+1.1
Factso3
- SimpleQA Verifiedo3 by 36.712.749.4+36.7
- AA-Omniscience · Accuracyo3 by 18.320.338.6+18.3
- AA-Omniscience · Non-hallucinationo3 by 4.67.311.9+4.6
Images and chartso3
- CharXiv (RQ)o3 by 21.856.878.6+21.8
- MathVistao3 by 13.773.186.8+13.7
- MMMU-Proo3 by 11.458.770.1+11.4
- MMMUo3 by 10.272.782.9+10.2
- LMArena · Visiono3 by 33 rating points11811214+33 rating
Following instructionso3
- IFBencho3 by 33.138.371.4+33.1
- Multi-Challengeo3 by 24.635.860.4+24.6
Other results16 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-1o3 by 57.33.560.8+57.3
- Aider-Polygloto3 by 46.634.781.3+46.6
- AIME 2025o3 by 46.240.286.4+46.2
- AA-Omniscienceo3 by 38.1-53.6-15.6+38.1
- AA Agentic Indexo3 by 34.31.836.1+34.3
- τ²-Bench Telecom (AA run)o3 by 27.852.980.7+27.8
- AIR-Bench 2024o3 by 24.160.484.5+24.1
- Artificial Analysis Coding Indexo3 by 18.220.238.4+18.2
- TAU-bench (retail)o3 by 18.155.873.9+18.1
- TAU-bench (airline)o3 by 163652+16
- HarmBencho3 by 12.885.698.4+12.8
- AA Intelligenceo3 by 1010.220.2+10
- bbqo3 by 5.892.197.9+5.8
- anthropic_red_teamGPT-4.1 mini by 199.398.3+1
- simple_safety_testsGPT-4.1 mini by 110099+1
- XSTesttie97.497.3tie
Questions people ask
Which is better, GPT-4.1 mini or o3?
o3 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. GPT-4.1 mini wins none, but costs 80% less.
Which is better for coding?
o3. It wins 2 of the 3 coding tests both models report; GPT-4.1 mini wins none, and 1 is a tie.
Which is cheaper?
GPT-4.1 mini costs $0.40 per million input tokens and $1.60 per million output tokens; o3 costs $2.00 and $8.00. That makes GPT-4.1 mini about 80% cheaper for the same work.
How do you compare the two?
We use the 36 benchmark tests both models have published scores on. The verdict counts the 20 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 16 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.