DeepSeek-R1 vs GPT-4o
Wins 4 of 6 areas
Coding · Reasoning · Long documents · Following instructions
Wins 1 of 6 areas
Facts
DeepSeek-R1 is the stronger all-rounder.GPT-4o is better at facts.
Scores updated · 52 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 thinking20DeepSeek-R12 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts20DeepSeek-R12 of 2 tests
- CodingWriting and fixing software21DeepSeek-R12 of 3 tests
- Following instructionsDoing exactly what it is asked10DeepSeek-R11 of 2 tests · 1 tie
- FactsGetting facts right instead of making them up12GPT-4o2 of 3 tests
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
Agents 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.
DeepSeek-R1 costs 70% 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 DeepSeek-R1 pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+18.2points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+16points ahead
- Questions about very long textsLongBench v2+10.2points ahead
Where GPT-4o pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+52.4points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+2.2points ahead
Every test, side by side
All 52 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-R1
- SWE-bench VerifiedDeepSeek-R1 by 1649.233.2+16
- SciCodeDeepSeek-R1 by 538.333.3+5
- Terminal-Bench HardGPT-4o by 2.26.18.3+2.2
ReasoningDeepSeek-R1
- GPQA DiamondDeepSeek-R1 by 18.270.852.6+18.2
- Humanity's Last ExamDeepSeek-R1 by 6.78.51.8+6.7
- CritPttie0.60tie
FactsGPT-4o
- AA-Omniscience · Non-hallucinationGPT-4o by 52.49.762.1+52.4
- AA-Omniscience · AccuracyDeepSeek-R1 by 6.830.523.7+6.8
- Vectara HHEM hallucination ratelower is betterGPT-4o by 1.711.39.6+1.7
Long documentsDeepSeek-R1
- LongBench v2DeepSeek-R1 by 10.258.348.1+10.2
- AA-LCRDeepSeek-R1 by 1.757.756+1.7
Following instructionsDeepSeek-R1
- IFBenchDeepSeek-R1 by 33936+3
- Multi-Challengetie40.740.3tie
Other results38 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.
- CodeforcesDeepSeek-R1 by 1270 rating points2029759+1270 rating
- Codeforces (Rating)DeepSeek-R1 by 1270 rating points2029759+1270 rating
- Codeforces (Percentile)DeepSeek-R1 by 72.796.323.6+72.7
- AIME 2024DeepSeek-R1 by 70.579.89.3+70.5
- CNMO 2024DeepSeek-R1 by 6878.810.8+68
- AIME 2025DeepSeek-R1 by 58.37011.7+58.3
- AlpacaEval2.0 (LC-winrate)DeepSeek-R1 by 36.587.651.1+36.5
- HarmBenchGPT-4o by 3547.982.9+35
- LiveCodeBench (Pass@1-COT)DeepSeek-R1 by 31.765.934.2+31.7
- LiveCodeBenchDeepSeek-R1 by 25.263.538.3+25.2
- MATH-500 (EM)DeepSeek-R1 by 22.797.374.6+22.7
- Aider-PolyglotDeepSeek-R1 by 22.653.330.7+22.6
- AA-OmniscienceGPT-4o by 21.7-32.2-10.5+21.7
- τ²-Bench Telecom (AA run)GPT-4o by 17.511.428.9+17.5
- MMLUDeepSeek-R1 by 1790.873.8+17
- C-EvalDeepSeek-R1 by 15.891.876+15.8
- MMLU-ProXDeepSeek-R1 by 14.475.561.1+14.4
- MATHDeepSeek-R1 by 1297.385.3+12
- Arena-Hard (GPT-4-1106 judge)DeepSeek-R1 by 11.992.380.4+11.9
- ARC-AGI-1DeepSeek-R1 by 11.315.84.5+11.3
- AIR-Bench 2024GPT-4o by 9.552.962.4+9.5
- MMLU-ProDeepSeek-R1 by 9.38474.7+9.3
- DROP (3-shot F1)DeepSeek-R1 by 8.592.283.7+8.5
- SimpleQAGPT-4o by 8.130.138.2+8.1
- vectara_avg_summary_lengthDeepSeek-R1 by 6.993.586.6+6.9
- Chinese SimpleQA (C-SimpleQA)DeepSeek-R1 by 563.758.7+5
- CLUEWSCDeepSeek-R1 by 4.992.887.9+4.9
- MMLU-ReduxDeepSeek-R1 by 4.992.988+4.9
- AA IntelligenceDeepSeek-R1 by 4.111.47.3+4.1
- vectara_answer_rateDeepSeek-R1 by 3.29793.8+3.2
- XSTestGPT-4o by 2.994.497.3+2.9
- IFEvalDeepSeek-R1 by 2.383.381+2.3
- FRAMES (Acc.)DeepSeek-R1 by 282.580.5+2
- anthropic_red_teamGPT-4o by 1.997.299.1+1.9
- vectara_factual_consistencyGPT-4o by 1.788.790.4+1.7
- bbqDeepSeek-R1 by 1.596.695.1+1.5
- simple_safety_teststie9898.5tie
- Artificial Analysis Coding Indextie24.624.2tie
Questions people ask
Which is better, DeepSeek-R1 or GPT-4o?
DeepSeek-R1 wins four of the six areas where both have results: coding, reasoning, long documents and following instructions. GPT-4o wins facts. They are level on agents.
Which is better for coding?
DeepSeek-R1. It wins 2 of the 3 coding tests both models report; GPT-4o wins 1.
Which is cheaper?
DeepSeek-R1 costs $2.00 per million input tokens and $4.00 per million output tokens; GPT-4o costs $5.00 and $15.00. That makes DeepSeek-R1 about 70% cheaper for the same work.
How do you compare the two?
We use the 52 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 38 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.