DeepSeek-V3 vs MiniMax M1 80K
Wins 2 of 6 areas
Coding · Facts
Wins 4 of 6 areas
Agents · Reasoning · Long documents · Following instructions
MiniMax M1 80K is the stronger all-rounder.DeepSeek-V3 is cheaper and better at facts.
Scores updated · 24 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 thinking02MiniMax M1 80K2 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts02MiniMax M1 80K2 of 2 tests
- Following instructionsDoing exactly what it is asked01MiniMax M1 80K1 of 2 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own01MiniMax M1 80K1 of 1 test
- FactsGetting facts right instead of making them up20DeepSeek-V32 of 2 tests
- CodingWriting and fixing software21DeepSeek-V32 of 3 tests
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-V3 costs 59% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where MiniMax M1 80K pulls ahead
- Reasons across sets of long documentsAA-LCR+17points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+14points ahead
- Keeps track of context across a multi-turn chatMulti-Challenge+13.3points ahead
Where DeepSeek-V3 pulls ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+12.2points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+5.6points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+3points ahead
Every test, side by side
All 24 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3
- SWE-bench VerifiedMiniMax M1 80K by 144256+14
- Terminal-Bench HardDeepSeek-V3 by 12.215.23+12.2
- SciCodeDeepSeek-V3 by 1.63937.4+1.6
ReasoningMiniMax M1 80K
- GPQA DiamondMiniMax M1 80K by 4.265.569.7+4.2
- Humanity's Last ExamMiniMax M1 80K by 4.24.78.9+4.2
- CritPttie00tie
FactsDeepSeek-V3
- AA-Omniscience · Non-hallucinationDeepSeek-V3 by 5.614.18.5+5.6
- AA-Omniscience · AccuracyDeepSeek-V3 by 325.422.5+3
Long documentsMiniMax M1 80K
- AA-LCRMiniMax M1 80K by 1740.757.7+17
- LongBench v2MiniMax M1 80K by 12.848.761.5+12.8
Following instructionsMiniMax M1 80K
- Multi-ChallengeMiniMax M1 80K by 13.331.444.7+13.3
- IFBenchtie4141.8tie
Other results11 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 2024MiniMax M1 80K by 46.839.286+46.8
- AIME 2025MiniMax M1 80K by 25.651.376.9+25.6
- LiveCodeBenchMiniMax M1 80K by 15.849.265+15.8
- τ²-Bench Telecom (AA run)DeepSeek-V3 by 12.947.134.2+12.9
- Artificial Analysis Coding IndexDeepSeek-V3 by 8.52314.5+8.5
- AA-OmniscienceDeepSeek-V3 by 7.8-40.7-48.5+7.8
- MATH-500 (EM)MiniMax M1 80K by 6.690.296.8+6.6
- SimpleQADeepSeek-V3 by 6.424.918.5+6.4
- MMLU-ProMiniMax M1 80K by 5.275.981.1+5.2
- ZebraLogicMiniMax M1 80K by 2.88486.8+2.8
- AA IntelligenceMiniMax M1 80K by 29.711.7+2
Questions people ask
Which is better, DeepSeek-V3 or MiniMax M1 80K?
MiniMax M1 80K wins four of the six areas where both have results: agents, reasoning, long documents and following instructions. DeepSeek-V3 wins coding and facts, and costs 59% less.
Which is better for coding?
DeepSeek-V3. It wins 2 of the 3 coding tests both models report; MiniMax M1 80K wins 1.
Which is cheaper?
DeepSeek-V3 costs $0.24 per million input tokens and $0.90 per million output tokens; MiniMax M1 80K costs $0.55 and $2.20. That makes DeepSeek-V3 about 59% cheaper for the same work.
How do you compare the two?
We use the 24 benchmark tests both models have published scores on. The verdict counts the 13 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 11 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.