o1 vs o3-mini
Wins 4 of 7 areas
Agents · Facts · Long documents · Following instructions
Wins 3 of 7 areas
Coding · Reasoning · Math
o1 wins more areas, narrowly.o3-mini is cheaper and better at reasoning.
Scores updated · 45 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.
- FactsGetting facts right instead of making them up30o13 of 3 tests
- AgentsCarrying out multi-step tasks on its own10o11 of 1 test
- Long documentsFinding answers in very long texts10o11 of 1 test
- Following instructionsDoing exactly what it is asked10o11 of 1 test
- ReasoningHard problems that need careful thinking12o3-mini2 of 4 tests · 1 tie
- CodingWriting and fixing software12o3-mini2 of 3 tests
- MathCompetition and research-level math01o3-mini1 of 1 test
Agents, math, long documents and following instructions 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.
o3-mini costs 93% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where o1 pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+25.8points ahead
- Reasons across sets of long documentsAA-LCR+22points ahead
- Common-sense trick questionsSimpleBench+17.3points ahead
Where o3-mini pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+8points ahead
- Code for real scientific research problemsSciCode+7points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+5points ahead
Every test, side by side
All 45 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Codingo3-mini
- SWE-bench Verifiedo3-mini by 841.349.3+8
- SciCodeo3-mini by 735.842.8+7
- Terminal-Bench Hardo1 by 6.812.96.1+6.8
Reasoningo3-mini
- SimpleBencho1 by 17.340.122.8+17.3
- Humanity's Last Examo3-mini by 5712+5
- GPQA Diamondo3-mini by 2.674.777.3+2.6
- CritPttie0.30.3tie
Factso1
- SimpleQA Verifiedo1 by 25.841.115.3+25.8
- AA-Omniscience · Accuracyo1 by 13.234.521.3+13.2
- AA-Omniscience · Non-hallucinationo1 by 11.530.418.9+11.5
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 2025o1 by 33.481.748.3+33.4
- LiveBencho3-mini by 32.352.384.6+32.3
- AA-Omniscienceo1 by 31.6-11.1-42.6+31.6
- τ²-Bench Telecom (AA run)o1 by 31.362.631.3+31.3
- SimpleQAo1 by 27.442.415+27.4
- Artificial Analysis Coding Indexo1 by 23.439.716.3+23.4
- TAU-bench (airline)o1 by 17.65032.4+17.6
- TAU-bench (retail)o1 by 13.270.857.6+13.2
- MMLU Language (0-shot) - Yorubao1 by 11.775.463.7+11.7
- MMLU Language (0-shot) - Swahilio1 by 11.685.473.8+11.6
- MMLU Language (0-shot) - Bengalio1 by 7.287.380.1+7.2
- MMLU Language (0-shot) - Hindio1 by 7.288.381.1+7.2
- MMLU Language (0-shot) - Averageo1 by 787.780.7+7
- MMLU Language (0-shot) - Italiano1 by 5.989.783.8+5.9
- MMLU Language (0-shot) - Indonesiano1 by 5.888.682.8+5.8
- MMLU Language (0-shot) - Japaneseo1 by 5.888.983.1+5.8
- MMLU Language (0-shot) - Frencho1 by 5.689.383.7+5.6
- MMLU Language (0-shot) - Koreano1 by 5.688.282.6+5.6
- MMLU Language (0-shot) - Portuguese (Brazil)o1 by 5.489.584.1+5.4
- MMLU Language (0-shot) - Chinese (Simplified)o1 by 5.388.983.6+5.3
- AIR-Bench 2024o1 by 5.18074.9+5.1
- Aider-Polygloto3-mini by 561.766.7+5
- AA Intelligenceo1 by 4.215.211+4.2
- FrontierMath (overall)o3-mini by 3.75.59.2+3.7
- bbqo1 by 3.297.394.1+3.2
- XSTesto1 by 2.99794.1+2.9
- MGSMo3-mini by 1.290.892+1.2
- HarmBencho1 by 1.196.395.2+1.1
- AIME 2024tie79.279.6tie
- anthropic_red_teamtie98.398.6tie
- simple_safety_teststie9999tie
Questions people ask
Which is better, o1 or o3-mini?
o1 wins four of the seven areas where both have results: agents, facts, long documents and following instructions. o3-mini wins coding, reasoning and math, and costs 93% less.
Which is better for coding?
o3-mini. It wins 2 of the 3 coding tests both models report; o1 wins 1.
Which is cheaper?
o1 costs $15.00 per million input tokens and $60.00 per million output tokens; o3-mini costs $1.10 and $4.40. That makes o3-mini about 93% cheaper for the same work.
How do you compare the two?
We use the 45 benchmark tests both models have published scores on. The verdict counts the 14 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.