DeepSeek-V3 vs GPT-4o mini
Wins 4 of 4 areas
Coding · Agents · Reasoning · Following instructions
Wins 0 of 4 areas
—
DeepSeek-V3 is the stronger all-rounder.GPT-4o mini is cheaper.
Scores updated · 32 tests both models report · How we compare
Where each one wins
Tests won in each of the four 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.
Following instructions 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.
GPT-4o mini costs 34% 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-V3 pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+33.3points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+22.9points ahead
- Code for real scientific research problemsSciCode+16.1points ahead
Where GPT-4o mini pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 32 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3
- SWE-bench VerifiedDeepSeek-V3 by 33.3428.7+33.3
- SciCodeDeepSeek-V3 by 16.13922.9+16.1
- Terminal-Bench 2.1DeepSeek-V3 by 11.316.95.6+11.3
ReasoningDeepSeek-V3
- GPQA DiamondDeepSeek-V3 by 22.965.542.6+22.9
- Humanity's Last Examtie4.74.2tie
Other results24 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.
- Aider-PolyglotDeepSeek-V3 by 4649.63.6+46
- HarmBenchGPT-4o mini by 35.249.784.9+35.2
- livecodebench_mediumDeepSeek-V3 by 28.353.525.2+28.3
- HumanEvalGPT-4o mini by 2265.287.2+22
- MMLUDeepSeek-V3 by 21.788.566.8+21.7
- AIR-Bench 2024GPT-4o mini by 15.540.856.3+15.5
- SimpleQADeepSeek-V3 by 1524.99.9+15
- MMLU-ProDeepSeek-V3 by 14.275.961.7+14.2
- LiveCodeBenchDeepSeek-V3 by 13.749.235.5+13.7
- DROPDeepSeek-V3 by 11.991.679.7+11.9
- Artificial Analysis Coding IndexDeepSeek-V3 by 11.62311.4+11.6
- MATHDeepSeek-V3 by 1090.280.2+10
- livecodebench_hardDeepSeek-V3 by 9.914.24.3+9.9
- GSM8KDeepSeek-V3 by 9.79484.3+9.7
- MBPPGPT-4o mini by 9.475.484.8+9.4
- naturalquestions_closedbookDeepSeek-V3 by 8.146.738.5+8.1
- MGSMGPT-4o mini by 7.279.887+7.2
- OpenBookQADeepSeek-V3 by 3.495.492+3.4
- AA IntelligenceDeepSeek-V3 by 39.76.7+3
- NarrativeQADeepSeek-V3 by 2.879.676.8+2.8
- simple_safety_testsGPT-4o mini by 2.595.397.8+2.5
- livecodebench_easyDeepSeek-V3 by 1.683.381.7+1.6
- anthropic_red_teamGPT-4o mini by 1.297.198.3+1.2
- XSTestDeepSeek-V3 by 1.197.196+1.1
Questions people ask
Which is better, DeepSeek-V3 or GPT-4o mini?
DeepSeek-V3 wins all four areas where both have results: coding, agents, reasoning and following instructions. GPT-4o mini wins none, but costs 34% less.
Which is better for coding?
DeepSeek-V3. It wins 3 of the 3 coding tests both models report; GPT-4o mini wins none.
Which is cheaper?
DeepSeek-V3 costs $0.24 per million input tokens and $0.90 per million output tokens; GPT-4o mini costs $0.15 and $0.60. That makes GPT-4o mini about 34% cheaper for the same work.
How do you compare the two?
We use the 32 benchmark tests both models have published scores on. The verdict counts the 8 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 24 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.