DeepSeek-V3 vs Mistral Large 3
Wins 4 of 6 areas
Coding · Facts · Long documents · Following instructions
Wins 2 of 6 areas
Agents · Reasoning
DeepSeek-V3 is the stronger all-rounder.Mistral Large 3 is better at reasoning.
Scores updated · 27 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 software20DeepSeek-V32 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up10DeepSeek-V31 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts10DeepSeek-V31 of 1 test
- Following instructionsDoing exactly what it is asked10DeepSeek-V31 of 1 test
- ReasoningHard problems that need careful thinking02Mistral Large 32 of 4 tests · 2 ties
- AgentsCarrying out multi-step tasks on its own01Mistral Large 31 of 3 tests · 2 ties
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.
DeepSeek-V3 costs 43% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where DeepSeek-V3 pulls ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+4.9points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+4.8points ahead
- Reasons across sets of long documentsAA-LCR+4.7points ahead
Where Mistral Large 3 pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+2.5points ahead
- Common-sense trick questionsSimpleBench+1.5points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+1.1points ahead
Every test, side by side
All 27 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3
- Terminal-Bench 2.1DeepSeek-V3 by 4.916.912+4.9
- SciCodeDeepSeek-V3 by 2.43936.6+2.4
- Terminal-Bench Hardtie15.215.9tie
AgentsMistral Large 3
- τ-Bench V3 · BankingMistral Large 3 by 1.14.75.8+1.1
- GDPValtie00tie
- Terminal-Bench 4.0tie00tie
ReasoningMistral Large 3
- GPQA DiamondMistral Large 3 by 2.565.568+2.5
- SimpleBenchMistral Large 3 by 1.518.920.4+1.5
- Humanity's Last Examtie4.74.2tie
- CritPttie00tie
FactsDeepSeek-V3
- Vectara HHEM hallucination ratelower is betterDeepSeek-V3 by 8.46.114.5+8.4
- AA-Omniscience · Accuracytie25.424.9tie
- AA-Omniscience · Non-hallucinationtie14.114tie
Following instructionsDeepSeek-V3
- IFBenchDeepSeek-V3 by 4.84136.2+4.8
Other results12 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.
- Arena HardDeepSeek-V3 by 36.391.455.1+36.3
- τ²-Bench Telecom (AA run)DeepSeek-V3 by 22.547.124.6+22.5
- LiveCodeBenchDeepSeek-V3 by 14.849.234.4+14.8
- vectara_factual_consistencyDeepSeek-V3 by 8.493.985.5+8.4
- MMLU-ReduxDeepSeek-V3 by 7.189.182+7.1
- MMMLUDeepSeek-V3 by 5.279.474.2+5.2
- vectara_avg_summary_lengthMistral Large 3 by 31 rating points81.7112.7+31 rating
- Artificial Analysis Coding IndexDeepSeek-V3 by 2.92320.1+2.9
- vectara_answer_rateMistral Large 3 by 1.397.598.8+1.3
- AA-OmniscienceMistral Large 3 by 1.1-40.7-39.6+1.1
- SimpleQADeepSeek-V3 by 1.124.923.8+1.1
- AA Intelligencetie9.79.3tie
Questions people ask
Which is better, DeepSeek-V3 or Mistral Large 3?
DeepSeek-V3 wins four of the six areas where both have results: coding, facts, long documents and following instructions. Mistral Large 3 wins agents and reasoning.
Which is better for coding?
DeepSeek-V3. It wins 2 of the 3 coding tests both models report; Mistral Large 3 wins none, and 1 is a tie.
Which is cheaper?
DeepSeek-V3 costs $0.24 per million input tokens and $0.90 per million output tokens; Mistral Large 3 costs $0.50 and $1.50. That makes DeepSeek-V3 about 43% cheaper for the same work.
How do you compare the two?
We use the 27 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 12 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.