DeepSeek-V3.2 vs MiniMax M2.5
Wins 2 of 6 areas
Reasoning · Facts
Wins 3 of 6 areas
Coding · Agents · Following instructions
MiniMax M2.5 wins more areas, narrowly.DeepSeek-V3.2 is cheaper and better at facts.
Scores updated · 28 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 software04MiniMax M2.54 of 5 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own02MiniMax M2.52 of 2 tests
- Following instructionsDoing exactly what it is asked01MiniMax M2.51 of 1 test
- FactsGetting facts right instead of making them up30DeepSeek-V3.23 of 3 tests
- ReasoningHard problems that need careful thinking20DeepSeek-V3.22 of 4 tests · 2 ties
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
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-V3.2 costs 53% 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 M2.5 pulls ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+39.8points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+24.9points ahead
- Real work tasks from 44 professionsGDPVal+24.1points ahead
Where DeepSeek-V3.2 pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+6.8points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+5.4points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+4.1points ahead
Every test, side by side
All 28 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingMiniMax M2.5
- SWE-bench ProMiniMax M2.5 by 39.815.655.4+39.8
- SWE-bench VerifiedMiniMax M2.5 by 7.173.180.2+7.1
- SciCodeMiniMax M2.5 by 3.738.942.6+3.7
- LMArena · WebDevMiniMax M2.5 by 25 rating points13621387+25 rating
- Terminal-Bench Hardtie35.634.8tie
AgentsMiniMax M2.5
- BrowseCompMiniMax M2.5 by 24.951.476.3+24.9
- GDPValMiniMax M2.5 by 24.19.833.9+24.1
ReasoningDeepSeek-V3.2
- Humanity's Last ExamDeepSeek-V3.2 by 4.124.620.5+4.1
- CritPtDeepSeek-V3.2 by 1.82.91.1+1.8
- ARC-AGI-2tie44.9tie
- GPQA Diamondtie8484.8tie
FactsDeepSeek-V3.2
- AA-Omniscience · AccuracyDeepSeek-V3.2 by 6.83326.2+6.8
- AA-Omniscience · Non-hallucinationDeepSeek-V3.2 by 5.417.311.9+5.4
- Vectara HHEM hallucination ratelower is betterDeepSeek-V3.2 by 2.86.39.1+2.8
Following instructionsMiniMax M2.5
- IFBenchMiniMax M2.5 by 10.960.771.6+10.9
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.
- AA Agentic IndexMiniMax M2.5 by 37.318.355.6+37.3
- Multi-SWE-BenchMiniMax M2.5 by 20.730.651.3+20.7
- AA-OmniscienceDeepSeek-V3.2 by 16.4-22.5-38.9+16.4
- swe_bench_bashMiniMax M2.5 by 15.86075.8+15.8
- vectara_avg_summary_lengthMiniMax M2.5 by 75.2 rating points62137.2+75.2 rating
- Artificial Analysis Coding IndexDeepSeek-V3.2 by 6.844.237.4+6.8
- ARC-AGI-1MiniMax M2.5 by 6.75763.7+6.7
- vectara_answer_rateMiniMax M2.5 by 5.692.698.2+5.6
- τ²-Bench Telecom (AA run)MiniMax M2.5 by 4.790.695.3+4.7
- AIME25 no toolsDeepSeek-V3.2 by 389.386.3+3
- vectara_factual_consistencyDeepSeek-V3.2 by 2.893.790.9+2.8
- AA IntelligenceMiniMax M2.5 by 1.321.522.8+1.3
Questions people ask
Which is better, DeepSeek-V3.2 or MiniMax M2.5?
MiniMax M2.5 wins three of the six areas where both have results: coding, agents and following instructions. DeepSeek-V3.2 wins reasoning and facts, and costs 53% less. They are level on long documents.
Which is better for coding?
MiniMax M2.5. It wins 4 of the 5 coding tests both models report; DeepSeek-V3.2 wins none, and 1 is a tie.
Which is cheaper?
DeepSeek-V3.2 costs $0.28 per million input tokens and $0.42 per million output tokens; MiniMax M2.5 costs $0.30 and $1.20. That makes DeepSeek-V3.2 about 53% cheaper for the same work.
How do you compare the two?
We use the 28 benchmark tests both models have published scores on. The verdict counts the 16 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.