DeepSeek-R1 vs o3-mini
Wins 2 of 6 areas
Agents · Long documents
Wins 2 of 6 areas
Reasoning · Following instructions
The two are evenly matched.DeepSeek-R1 is better at agents and long documents; o3-mini at reasoning and following instructions.
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.
- ReasoningHard problems that need careful thinking02o3-mini2 of 3 tests · 1 tie
- Following instructionsDoing exactly what it is asked01o3-mini1 of 2 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own10DeepSeek-R11 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts10DeepSeek-R11 of 1 test
- CodingWriting and fixing software11Even1 each · 2 ties
- FactsGetting facts right instead of making them up11Even1 each
Long documents rests on a single test.
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 8% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where o3-mini pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+28.1points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+9.2points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+6.5points ahead
Where DeepSeek-R1 pulls ahead
- Reasons across sets of long documentsAA-LCR+14.7points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+14.6points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+9.2points 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.
CodingEven
- Terminal-Bench 2.1DeepSeek-R1 by 14.619.14.5+14.6
- SciCodeo3-mini by 4.538.342.8+4.5
- SWE-bench Verifiedtie49.249.3tie
- Terminal-Bench Hardtie6.16.1tie
AgentsDeepSeek-R1
- τ-Bench V3 · BankingDeepSeek-R1 by 1.26.45.2+1.2
- GDPValtie00tie
- Terminal-Bench 4.0tie00tie
Reasoningo3-mini
- GPQA Diamondo3-mini by 6.570.877.3+6.5
- Humanity's Last Examo3-mini by 3.58.512+3.5
- CritPttie0.60.3tie
FactsEven
- AA-Omniscience · AccuracyDeepSeek-R1 by 9.230.521.3+9.2
- AA-Omniscience · Non-hallucinationo3-mini by 9.29.718.9+9.2
Following instructionso3-mini
- IFBencho3-mini by 28.13967.1+28.1
- Multi-Challengetie40.739.9tie
Other results19 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.
- HarmBencho3-mini by 47.347.995.2+47.3
- AIR-Bench 2024o3-mini by 2252.974.9+22
- AIME 2025DeepSeek-R1 by 21.77048.3+21.7
- τ²-Bench Telecom (AA run)o3-mini by 19.911.431.3+19.9
- ARC-AGI-1o3-mini by 18.715.834.5+18.7
- SimpleQADeepSeek-R1 by 15.130.115+15.1
- LiveCodeBencho3-mini by 14.263.577.7+14.2
- Aider-Polygloto3-mini by 13.453.366.7+13.4
- HMMT Feb. 2025DeepSeek-R1 by 13.441.728.3+13.4
- IFEvalo3-mini by 10.683.393.9+10.6
- AA-OmniscienceDeepSeek-R1 by 10.4-32.2-42.6+10.4
- Artificial Analysis Coding IndexDeepSeek-R1 by 8.324.616.3+8.3
- usamo_2025DeepSeek-R1 by 2.74.82.1+2.7
- bbqDeepSeek-R1 by 2.596.694.1+2.5
- anthropic_red_teamo3-mini by 1.497.298.6+1.4
- simple_safety_testso3-mini by 19899+1
- AA Intelligencetie11.411tie
- XSTesttie94.494.1tie
- AIME 2024tie79.879.6tie
Questions people ask
Which is better, DeepSeek-R1 or o3-mini?
DeepSeek-R1 and o3-mini each win two of the six areas where both have results. DeepSeek-R1 wins agents and long documents; o3-mini wins reasoning and following instructions. o3-mini costs 8% less. They are level on coding and facts.
Which is better for coding?
Neither. They win 1 coding test each of the 4 both models report, and 2 are ties.
Which is cheaper?
DeepSeek-R1 costs $2.00 per million input tokens and $4.00 per million output tokens; o3-mini costs $1.10 and $4.40. That makes o3-mini about 8% 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 15 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 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.