DeepSeek-R1 0528 vs GPT-4o
Wins 6 of 6 areas
Coding · Agents · Reasoning · Facts · Long documents · Following instructions
Wins 0 of 6 areas
—
DeepSeek-R1 0528 is the stronger all-rounder.
Scores updated · 39 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.
- CodingWriting and fixing software30DeepSeek-R1 05283 of 3 tests
- ReasoningHard problems that need careful thinking30DeepSeek-R1 05283 of 3 tests
- Following instructionsDoing exactly what it is asked20DeepSeek-R1 05282 of 2 tests
- FactsGetting facts right instead of making them up21DeepSeek-R1 05282 of 3 tests
- Long documentsFinding answers in very long texts10DeepSeek-R1 05281 of 2 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own10DeepSeek-R1 05281 of 1 test
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 0528 costs 78% 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 0528 pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+28.7points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+14points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+11.4points ahead
Where GPT-4o pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+45.5points ahead
Every test, side by side
All 39 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-R1 0528
- SWE-bench VerifiedDeepSeek-R1 0528 by 11.444.633.2+11.4
- Terminal-Bench HardDeepSeek-R1 0528 by 7.615.98.3+7.6
- SciCodeDeepSeek-R1 0528 by 740.333.3+7
ReasoningDeepSeek-R1 0528
- GPQA DiamondDeepSeek-R1 0528 by 28.781.352.6+28.7
- Humanity's Last ExamDeepSeek-R1 0528 by 1415.81.8+14
- CritPtDeepSeek-R1 0528 by 1.41.40+1.4
FactsDeepSeek-R1 0528
- AA-Omniscience · Non-hallucinationGPT-4o by 45.516.662.1+45.5
- AA-Omniscience · AccuracyDeepSeek-R1 0528 by 6.830.523.7+6.8
- SimpleQA VerifiedDeepSeek-R1 0528 by 1.427.426+1.4
Long documentsDeepSeek-R1 0528
- LongBench v2DeepSeek-R1 0528 by 452.148.1+4
- AA-LCRtie55.756tie
Following instructionsDeepSeek-R1 0528
- Multi-ChallengeDeepSeek-R1 0528 by 4.74540.3+4.7
- IFBenchDeepSeek-R1 0528 by 3.639.636+3.6
Other results25 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 2024DeepSeek-R1 0528 by 82.191.49.3+82.1
- AIME 2025DeepSeek-R1 0528 by 75.887.511.7+75.8
- livecodebench_hardDeepSeek-R1 0528 by 59.163.64.5+59.1
- livecodebench_mediumDeepSeek-R1 0528 by 58.890.932.1+58.8
- SimpleQADeepSeek-R1 0528 by 54.192.338.2+54.1
- Aider-PolyglotDeepSeek-R1 0528 by 40.971.630.7+40.9
- LiveCodeBenchDeepSeek-R1 0528 by 3573.338.3+35
- HarmBenchGPT-4o by 28.354.682.9+28.3
- MATH-500 (EM)DeepSeek-R1 0528 by 23.49874.6+23.4
- AA-OmniscienceGPT-4o by 16.9-27.4-10.5+16.9
- ARC-AGI-1DeepSeek-R1 0528 by 16.721.24.5+16.7
- livecodebench_easyDeepSeek-R1 0528 by 16.799.282.5+16.7
- AA Agentic IndexDeepSeek-R1 0528 by 12.420.88.4+12.4
- TAU-bench (airline)DeepSeek-R1 0528 by 10.753.542.8+10.7
- MMLU-ProDeepSeek-R1 0528 by 10.38574.7+10.3
- τ²-Bench Telecom (AA run)DeepSeek-R1 0528 by 7.636.528.9+7.6
- AA IntelligenceDeepSeek-R1 0528 by 5.813.17.3+5.8
- MMLU-ReduxDeepSeek-R1 0528 by 5.493.488+5.4
- TAU-bench (retail)DeepSeek-R1 0528 by 3.663.960.3+3.6
- FRAMES (Acc.)DeepSeek-R1 0528 by 2.58380.5+2.5
- XSTestDeepSeek-R1 0528 by 1.598.897.3+1.5
- bbqDeepSeek-R1 0528 by 1.496.595.1+1.4
- Artificial Analysis Coding Indextie2424.2tie
- simple_safety_teststie98.398.5tie
- anthropic_red_teamtie99.199.1tie
Questions people ask
Which is better, DeepSeek-R1 0528 or GPT-4o?
DeepSeek-R1 0528 wins all six areas where both have results: coding, agents, reasoning, facts, long documents and following instructions. GPT-4o wins none.
Which is better for coding?
DeepSeek-R1 0528. It wins 3 of the 3 coding tests both models report; GPT-4o wins none.
Which is cheaper?
DeepSeek-R1 0528 costs $1.35 per million input tokens and $3.00 per million output tokens; GPT-4o costs $5.00 and $15.00. That makes DeepSeek-R1 0528 about 78% cheaper for the same work.
How do you compare the two?
We use the 39 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 25 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.