Mistral Medium 3 vs MiMo-V2.5
Updated October 21, 2026. 50 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Only MiMo-V2.5 has a coding score, 50.0 on the Vibe Coding Index; Mistral Medium 3 has none, so the two cannot be ranked on coding.
Mistral Medium 3 has the edge where the benchmark that matters is LMArena Text: Japanese, where it leads.
- Shared results
- 50
- Mistral Medium 3 only
- 7
- MiMo-V2.5 only
- 30
- Like-for-like categories
- 8 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Only MiMo-V2.5 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
MiMo-V2.5
MiMo-V2.5 leads, 21.8 against 9.0.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only MiMo-V2.5 has a AA Agentic Index result, 15.8.
Directional only
Reasoning
GPQA Diamond
No call
Only MiMo-V2.5 has a GPQA Diamond result, 81.6%.
Directional only
Knowledge
MMLU Pro
No call
Only MiMo-V2.5 has a MMLU Pro result, 82.9%.
Directional only
Web design
Design Arena: websites
MiMo-V2.5
MiMo-V2.5 leads, 1266 against 1079.
Same board, both measured
Long documents
Listed context window
MiMo-V2.5
MiMo-V2.5 takes 1.05M tokens against 131K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
MiMo-V2.5
Mistral Medium 3 $0.0014, MiMo-V2.5 $0.0003 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only MiMo-V2.5 has the number, $0.01.
Directional only
At a glance
- Coding index
- Mistral Medium 3 not benchmarked, MiMo-V2.5 50.0 from 3 benchmarks.
- Widest category
- Agentic work, 66.4 points apart on average over 1 shared benchmark.
- Widest single gap
- τ²-Bench Telecom: Mistral Medium 3 24.3% against MiMo-V2.5 90.6%, 66.4 points apart on a 0 to 100 scale.
- Output price
- Mistral Medium 3 $2.00 against MiMo-V2.5 $0.336 per million output tokens, 6.0x apart.
- Context window
- Mistral Medium 3 takes 131K tokens, MiMo-V2.5 1.05M.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Mistral Medium 3 | Bars | MiMo-V2.5 | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 0 | 7 | ||
| Math | 0 | 2 | ||
| Knowledge | 0 | 4 | ||
| Web design | 0 | 6 | ||
| General | 1 | 25 | ||
| All shared boards | 1 | 49 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index03
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| LMArena WebDev | Not measured | 1438 | Only MiMo-V2.5 measured |
| AA Coding Index | Not measured | 56.8 | Only MiMo-V2.5 measured |
| LiveCodeBench (Vals AI) | Not measured | 81.5% | Only MiMo-V2.5 measured |
| SWE-bench Verified (Vals AI) | Not measured | 71.0% | Only MiMo-V2.5 measured |
| Vibe Code Bench (Vals AI) | Not measured | 42.2% | Only MiMo-V2.5 measured |
| Code migration (Vals AI) | Not measured | 14.2% | Only MiMo-V2.5 measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 60.7% | Only MiMo-V2.5 measured |
| LiveCodeBench (Artificial Analysis run) | 40.0% | Not measured | Only Mistral Medium 3 measured |
| SciCode | Not measured | 43.9% | Only MiMo-V2.5 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 3.8% | 41.7% | MiMo-V2.5 |
| ALE-Bench | Not measured | 514.0 | Only MiMo-V2.5 measured |
| LMArena Text: coding | 1385 | 1469 | MiMo-V2.5 |
| LMArena Text: software and IT | 1385 | 1463 | MiMo-V2.5 |
| LMArena WebDev: React | Not measured | 1425 | Only MiMo-V2.5 measured |
| LMArena WebDev: HTML | Not measured | 1429 | Only MiMo-V2.5 measured |
| Neither model has a result on the other 23 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| AA Intelligence Index | 9.0 | 21.8 | MiMo-V2.5 |
| Epoch Capabilities Index | 134.0 | Not measured | Only Mistral Medium 3 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| AA Agentic Index | Not measured | 15.8 | Only MiMo-V2.5 measured |
| Finance Agent (Vals AI) | Not measured | 36.7% | Only MiMo-V2.5 measured |
| Tax Agent Bench (Vals AI) | Not measured | 7.8% | Only MiMo-V2.5 measured |
| Legal Research Bench (Vals AI) | Not measured | 9.1% | Only MiMo-V2.5 measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 55.1% | Only MiMo-V2.5 measured |
| τ²-Bench Telecom | 24.3% | 90.6% | MiMo-V2.5 |
| Neither model has a result on the other 33 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index07
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 81.6% | Only MiMo-V2.5 measured |
| MMMU Pro (Vals AI) | Not measured | 80.0% | Only MiMo-V2.5 measured |
| GPQA Diamond (Artificial Analysis run) | 57.8% | 84.9% | MiMo-V2.5 |
| Humanity's Last Exam (Artificial Analysis run) | 4.1% | 27.2% | MiMo-V2.5 |
| LMArena Text: hard prompts | 1364 | 1450 | MiMo-V2.5 |
| LMArena Text: expert | 1342 | 1459 | MiMo-V2.5 |
| LMArena Text: hard prompts in English | 1374 | 1461 | MiMo-V2.5 |
| LMArena Vision: diagrams | 1155 | 1269 | MiMo-V2.5 |
| LMArena Vision: homework | 1165 | 1262 | MiMo-V2.5 |
| Neither model has a result on the other 12 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| AA Math Index | 30.3 | Not measured | Only Mistral Medium 3 measured |
| AIME 2025 (Artificial Analysis run) | 30.3% | Not measured | Only Mistral Medium 3 measured |
| ProofBench | Not measured | 16.0% | Only MiMo-V2.5 measured |
| LMArena Text: math | 1353 | 1436 | MiMo-V2.5 |
| LMArena Text: mathematical work | 1354 | 1451 | MiMo-V2.5 |
| Neither model has a result on the other 8 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 82.9% | Only MiMo-V2.5 measured |
| LegalBench (Vals AI) | Not measured | 78.9% | Only MiMo-V2.5 measured |
| CorpFin (Vals AI) | Not measured | 59.9% | Only MiMo-V2.5 measured |
| TaxEval (Vals AI) | Not measured | 71.8% | Only MiMo-V2.5 measured |
| MortgageTax (Vals AI) | Not measured | 59.3% | Only MiMo-V2.5 measured |
| MedCode (Vals AI) | Not measured | 31.9% | Only MiMo-V2.5 measured |
| MMLU Pro (Artificial Analysis run) | 76.0% | Not measured | Only Mistral Medium 3 measured |
| LMArena Text: science | 1381 | 1440 | MiMo-V2.5 |
| LMArena Text: medicine and healthcare | 1381 | 1432 | MiMo-V2.5 |
| LMArena Text: legal and government | 1384 | 1425 | MiMo-V2.5 |
| LMArena Text: business and finance | 1352 | 1432 | MiMo-V2.5 |
| Neither model has a result on the other 6 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index06
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| Design Arena: websites | 1079 | 1266 | MiMo-V2.5 |
| Design Arena: UI components | 1024 | 1258 | MiMo-V2.5 |
| Design Arena: game dev | 1018 | 1249 | MiMo-V2.5 |
| Design Arena: data viz | 1034 | 1249 | MiMo-V2.5 |
| Design Arena: 3D | 1083 | 1216 | MiMo-V2.5 |
| Design Arena: all code categories | 1069 | 1260 | MiMo-V2.5 |
| Design Arena: SVG | Not measured | 1179 | Only MiMo-V2.5 measured |
| Design Arena: ASCII art | Not measured | 1141 | Only MiMo-V2.5 measured |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index125
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| SAGE (Vals AI) | Not measured | 43.3% | Only MiMo-V2.5 measured |
| MedScribe (Vals AI) | Not measured | 72.2% | Only MiMo-V2.5 measured |
| Vals Multimodal Index | Not measured | 52.8% | Only MiMo-V2.5 measured |
| IFBench | 39.3% | 67.1% | MiMo-V2.5 |
| Long context reasoning (AA-LCR) | 31.3% | 73.0% | MiMo-V2.5 |
| DTBench | 62.3% | Not measured | Only Mistral Medium 3 measured |
| LMArena Text | 1370 | 1428 | MiMo-V2.5 |
| LMArena Text: instruction following | 1338 | 1435 | MiMo-V2.5 |
| LMArena Text: creative writing | 1343 | 1393 | MiMo-V2.5 |
| LMArena Vision | 1157 | 1247 | MiMo-V2.5 |
| LMArena Text: English | 1385 | 1447 | MiMo-V2.5 |
| LMArena Text: non-English | 1353 | 1405 | MiMo-V2.5 |
| LMArena Text: Chinese | 1374 | 1469 | MiMo-V2.5 |
| LMArena Text: German | 1377 | 1421 | MiMo-V2.5 |
| LMArena Text: Japanese | 1316 | 1306 | Mistral Medium 3 |
| LMArena Text: Korean | 1304 | 1361 | MiMo-V2.5 |
| LMArena Text: French | 1368 | 1448 | MiMo-V2.5 |
| LMArena Text: Spanish | 1364 | 1414 | MiMo-V2.5 |
| LMArena Text: Russian | 1359 | 1396 | MiMo-V2.5 |
| LMArena Text: Polish | 1371 | 1413 | MiMo-V2.5 |
| LMArena Text: multi-turn | 1384 | 1445 | MiMo-V2.5 |
| LMArena Text: longer queries | 1360 | 1445 | MiMo-V2.5 |
| LMArena Text: writing and language | 1352 | 1409 | MiMo-V2.5 |
| LMArena Text: entertainment and media | 1335 | 1399 | MiMo-V2.5 |
| LMArena Vision: English | 1172 | 1259 | MiMo-V2.5 |
| LMArena Vision: Chinese | 1161 | 1326 | MiMo-V2.5 |
| LMArena Vision: reading text | 1154 | 1256 | MiMo-V2.5 |
| LMArena Vision: humor | 1145 | 1226 | MiMo-V2.5 |
| LMArena Vision: captioning | 1094 | Not measured | Only Mistral Medium 3 measured |
| LMArena Vision: recognizing things | 1141 | 1189 | MiMo-V2.5 |
| LMArena Vision: creative writing | 1146 | 1230 | MiMo-V2.5 |
| Neither model has a result on the other 15 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Mistral Medium 3 | MiMo-V2.5 | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
MiMo-V2.5 costs less
Repository review
50K input + 3K output tokens
MiMo-V2.5 costs less
Agent loop
200K input + 10K output tokens
MiMo-V2.5 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | Mistral Medium 3 | Xiaomi MiMo-V2.5 |
|---|---|---|
| Maker | Mistral AI | Xiaomi |
| Weights | No public weights listed | Open weights |
| Released | May 7, 2025 | April 22, 2026 |
| Context window | 131K | 1.05M |
| Input $/Mtok | $0.40 | $0.168 |
| Output $/Mtok | $2.00 | $0.336 |
| Coding index | Not benchmarked | 50.0 |
| Independent measurements behind it | 0 | 3 |
| Cost per solved task | Not computable | $0.01 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Mistral Medium 3 better than MiMo-V2.5 for coding?
Only MiMo-V2.5 has a coding score, 50.0 on the Vibe Coding Index; Mistral Medium 3 has none, so the two cannot be ranked on coding. Mistral Medium 3 has the edge where the benchmark that matters is LMArena Text: Japanese, where it leads.
Which is cheaper, Mistral Medium 3 or MiMo-V2.5?
Mistral Medium 3 costs $0.40 per million input tokens and $2.00 per million output tokens; MiMo-V2.5 costs $0.168 and $0.336. Cost per solved task is known for MiMo-V2.5 only.
Which has the bigger context window, Mistral Medium 3 or MiMo-V2.5?
MiMo-V2.5, with 1.05M tokens against 131K.
How many benchmarks is the Mistral Medium 3 vs MiMo-V2.5 comparison based on?
50 benchmarks have a result for both models as of October 21, 2026. Mistral Medium 3's coding index rests on 0 independent benchmarks and MiMo-V2.5's on 3. Where only one model has a result, the page says so instead of naming a winner.