MiMo V2.5 vs Qwen3.5 397B A17B
Wins 1 of 8 areas
Coding
Wins 5 of 8 areas
Reasoning · Facts · Images and charts · Long documents · Following instructions
Qwen3.5 397B A17B is the stronger all-rounder.MiMo V2.5 is cheaper and better at coding.
Scores updated · 34 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- Images and chartsUnderstanding pictures, charts and video02Qwen3.5 397B A17B2 of 2 tests
- ReasoningHard problems that need careful thinking12Qwen3.5 397B A17B2 of 3 tests
- FactsGetting facts right instead of making them up12Qwen3.5 397B A17B2 of 3 tests
- Long documentsFinding answers in very long texts01Qwen3.5 397B A17B1 of 1 test
- Following instructionsDoing exactly what it is asked01Qwen3.5 397B A17B1 of 1 test
- CodingWriting and fixing software32MiMo V2.53 of 7 tests · 2 ties
- AgentsCarrying out multi-step tasks on its own11Even1 each · 1 tie
- MathCompetition and research-level math11Even1 each
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.
MiMo V2.5 costs 92% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Qwen3.5 397B A17B pulls ahead
- Recent programming contest problemsLiveCodeBench v6+48.1points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+45.6points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+14points ahead
Where MiMo V2.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+57points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+12.4points ahead
- Real work tasks from 44 professionsGDPVal+10.3points ahead
Every test, side by side
All 34 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingMiMo V2.5
- LiveCodeBench v6Qwen3.5 397B A17B by 48.135.583.6+48.1
- SWE-bench VerifiedQwen3.5 397B A17B by 45.630.876.4+45.6
- Terminal-Bench 2.1MiMo V2.5 by 12.463.751.3+12.4
- SWE-bench ProMiMo V2.5 by 5.256.150.9+5.2
- LMArena · WebDevMiMo V2.5 by 38 rating points14381400+38 rating
- SciCodetie43.944.8tie
- Terminal-Bench Hardtie41.740.9tie
AgentsEven
- GDPValMiMo V2.5 by 10.324.914.6+10.3
- τ-Bench V3 · BankingQwen3.5 397B A17B by 4.78.713.4+4.7
- Terminal-Bench 4.0tie00tie
ReasoningQwen3.5 397B A17B
- GPQA DiamondQwen3.5 397B A17B by 4.484.989.3+4.4
- CritPtMiMo V2.5 by 23.71.7+2
- Humanity's Last ExamQwen3.5 397B A17B by 1.827.229+1.8
FactsQwen3.5 397B A17B
- AA-Omniscience · Non-hallucinationMiMo V2.5 by 5768.111.1+57
- AA-Omniscience · AccuracyQwen3.5 397B A17B by 1416.830.8+14
- SimpleQA VerifiedQwen3.5 397B A17B by 9.916.126+9.9
Images and chartsQwen3.5 397B A17B
- MMMU-ProQwen3.5 397B A17B by 1.975.477.3+1.9
- LMArena · VisionQwen3.5 397B A17B by 14 rating points12481262+14 rating
MathEven
- HMMT Feb. 2026Qwen3.5 397B A17B by 5.382.687.9+5.3
- AIME 2026MiMo V2.5 by 2.393.691.3+2.3
Following instructionsQwen3.5 397B A17B
- IFBenchQwen3.5 397B A17B by 11.767.178.8+11.7
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.
- GDPval-AA v2MiMo V2.5 by 303 rating points1265962+303 rating
- MMLU-ProQwen3.5 397B A17B by 2265.887.8+22
- AA-OmniscienceMiMo V2.5 by 20.9-9.8-30.8+20.9
- Terminal-Bench 2.0MiMo V2.5 by 13.365.852.5+13.3
- Toolathlon VerifiedMiMo V2.5 by 8.449.140.7+8.4
- HLE (with tools)Qwen3.5 397B A17B by 8.34048.3+8.3
- SWEBench Pro PublicMiMo V2.5 by 5.256.150.9+5.2
- MMLU-ReduxQwen3.5 397B A17B by 5.189.894.9+5.1
- τ²-Bench Telecom (AA run)Qwen3.5 397B A17B by 590.695.6+5
- τ³-Bench BankingQwen3.5 397B A17B by 4.4913.4+4.4
- AA IntelligenceMiMo V2.5 by 3.421.818.4+3.4
- StrongREJECTtie99.399.4tie
Questions people ask
Which is better, MiMo V2.5 or Qwen3.5 397B A17B?
Qwen3.5 397B A17B wins five of the eight areas we test: reasoning, facts, images and charts, long documents and following instructions. MiMo V2.5 wins coding, and costs 92% less. They are level on agents and math.
Which is better for coding?
MiMo V2.5. It wins 3 of the 7 coding tests both models report; Qwen3.5 397B A17B wins 2, and 2 are ties.
Which is cheaper?
MiMo V2.5 costs $0.11 per million input tokens and $0.22 per million output tokens; Qwen3.5 397B A17B costs $0.60 and $3.60. That makes MiMo V2.5 about 92% cheaper for the same work.
How do you compare the two?
We use the 34 benchmark tests both models have published scores on. The verdict counts the 22 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.