o1 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.
Scores updated · 54 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.
- ReasoningHard problems that need careful thinking03o33 of 4 tests · 1 tie
- CodingWriting and fixing software03o33 of 3 tests
- Images and chartsUnderstanding pictures, charts and video03o33 of 3 tests
- FactsGetting facts right instead of making them up12o32 of 3 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
- Following instructionsDoing exactly what it is asked01o31 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 costs 87% 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+27.8points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+24.2points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+18.6points ahead
Where o1 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+18.5points ahead
Every test, side by side
All 54 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Codingo3
- SWE-bench Verifiedo3 by 27.841.369.1+27.8
- Terminal-Bench Hardo3 by 24.212.937.1+24.2
- SciCodeo3 by 5.235.841+5.2
Reasoningo3
- Humanity's Last Examo3 by 13.1720.1+13.1
- SimpleBencho3 by 1340.153.1+13
- GPQA Diamondo3 by 874.782.7+8
- CritPttie0.31.1tie
Factso3
- AA-Omniscience · Non-hallucinationo1 by 18.530.411.9+18.5
- SimpleQA Verifiedo3 by 8.341.149.4+8.3
- AA-Omniscience · Accuracyo3 by 4.134.538.6+4.1
Images and chartso3
- MathVistao3 by 1571.886.8+15
- MMMUo3 by 5.377.682.9+5.3
- LMArena · Visiono3 by 46 rating points11681214+46 rating
Other results37 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.
- metr_hcasto3 by 22.142.664.7+22.1
- Aider-Polygloto3 by 19.661.781.3+19.6
- τ²-Bench Telecom (AA run)o3 by 18.162.680.7+18.1
- metr_re_bencho3 by 15015+15
- AIME 2024o3 by 12.479.291.6+12.4
- FrontierMath (overall)o3 by 10.35.515.8+10.3
- SimpleQAo3 by 742.449.4+7
- AA Agentic Indexo3 by 531.136.1+5
- AA Intelligenceo3 by 515.220.2+5
- AIME 2025o3 by 4.781.786.4+4.7
- AA-Omniscienceo1 by 4.5-11.1-15.6+4.5
- AIR-Bench 2024o3 by 4.58084.5+4.5
- TAU-bench (retail)o3 by 3.170.873.9+3.1
- HarmBencho3 by 2.196.398.4+2.1
- TAU-bench (airline)o3 by 25052+2
- MATH-500 (EM)o3 by 1.796.498.1+1.7
- MMLU Language (0-shot) - Hindio3 by 1.588.389.8+1.5
- MMLU Language (0-shot) - Italiano3 by 1.589.791.2+1.5
- Artificial Analysis Coding Indexo1 by 1.339.738.4+1.3
- MMLU Language (0-shot) - Frencho3 by 1.389.390.6+1.3
- MMLU Language (0-shot) - Indonesiano3 by 1.288.689.8+1.2
- MMLU Language (0-shot) - Spanisho3 by 1.289.991.1+1.2
- MMLU Language (0-shot) - Averageo3 by 1.187.788.8+1.1
- MMLU Language (0-shot) - Koreano3 by 1.188.289.3+1.1
- bbqtie97.397.9tie
- MMLU Language (0-shot) - Bengalitie87.387.8tie
- MMLU Language (0-shot) - Chinese (Simplified)tie88.989.3tie
- XSTesttie9797.3tie
- BBQ P(not stereotyping | ambiguous question, not unknown)tie0.10.3tie
- BBQ Accuracy on Ambiguous Questionstie10.9tie
- PersonQA hallucination ratelower is bettertie0.20.3tie
- SimpleQA hallucination ratelower is bettertie0.40.5tie
- anthropic_red_teamtie98.398.3tie
- BBQ Accuracy on Unambiguous Questionstie0.90.9tie
- simple_safety_teststie9999tie
- SimpleQA accuracytie0.50.5tie
- StrongREJECTtie9797tie
Questions people ask
Which is better, o1 or o3?
o3 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. o1 wins none.
Which is better for coding?
o3. It wins 3 of the 3 coding tests both models report; o1 wins none.
Which is cheaper?
o1 costs $15.00 per million input tokens and $60.00 per million output tokens; o3 costs $2.00 and $8.00. That makes o3 about 87% cheaper for the same work.
How do you compare the two?
We use the 54 benchmark tests both models have published scores on. The verdict counts the 17 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 37 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.