Mistral Large 4 vs Qwen2.5 72B Instruct
Mistral Large 4 leads Qwen2.5 72B Instruct on the Artificial Analysis Intelligence Index, 38.4 to 7.7, and Qwen2.5 72B Instruct costs 10.4x less per output token ($0.40 vs $4.18 per million). Updated October 21, 2026; 32 benchmarks have a result for both.
Mistral Large 4 or Qwen2.5 72B Instruct: which to pick
Only Mistral Large 4 has a coding score, 56.4 on the Vibe Coding Index; Qwen2.5 72B Instruct has none, so the two cannot be ranked on coding.
Pick Qwen2.5 72B Instruct for output tokens 10.4x cheaper ($0.40 vs $4.18 per million).
- Shared results
- 32
- Mistral Large 4 only
- 49
- Qwen2.5 72B Instruct only
- 21
- Like-for-like categories
- 6 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
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 Mistral Large 4 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
Mistral Large 4
Mistral Large 4 leads, 38.4 against 7.7.
Same board, both measured
Web design
Design Arena: websites
No call
Only Mistral Large 4 has a Design Arena: websites result, 1258.
Directional only
Long documents
Listed context window
Mistral Large 4
Mistral Large 4 takes 1.05M tokens against 33K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen2.5 72B Instruct
Mistral Large 4 $0.0034, Qwen2.5 72B Instruct $0.0006 per turn.
List prices
At a glance
- Coding index
- Mistral Large 4 56.4 from 5 benchmarks, Qwen2.5 72B Instruct not benchmarked.
- Widest category
- Math, 50.0 points apart on average over 3 shared benchmarks.
- Widest single gap
- OTIS Mock AIME: Mistral Large 4 98.3% against Qwen2.5 72B Instruct 8.1%, 90.3 points apart on a 0 to 100 scale.
- Output price
- Mistral Large 4 $4.18 against Qwen2.5 72B Instruct $0.40 per million output tokens, 10.4x apart.
- Context window
- Mistral Large 4 takes 1.05M tokens, Qwen2.5 72B Instruct 33K.
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 Large 4 | Bars | Qwen2.5 72B Instruct | Ties |
|---|---|---|---|---|
| Coding | 2 | 0 | ||
| Intelligence | 2 | 0 | ||
| Reasoning | 5 | 0 | ||
| Math | 3 | 0 | ||
| Knowledge | 4 | 0 | ||
| General | 16 | 0 | ||
| All shared boards | 32 | 0 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index20
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| LiveBench coding | 60.0% | Not measured | Only Mistral Large 4 measured |
| LMArena WebDev | 1543 | Not measured | Only Mistral Large 4 measured |
| Terminal-Bench 4 (Vals AI) | 27.8% | Not measured | Only Mistral Large 4 measured |
| Vibe Code Bench (Vals AI) | 78.4% | Not measured | Only Mistral Large 4 measured |
| Code migration (Vals AI) | 30.6% | Not measured | Only Mistral Large 4 measured |
| IOI (Vals AI) | 45.3% | Not measured | Only Mistral Large 4 measured |
| SWE-bench Multimodal (Vals AI) | 33.1% | Not measured | Only Mistral Large 4 measured |
| Vibe Code Bench 1-100 (Vals AI) | 14.3% | Not measured | Only Mistral Large 4 measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 27.6% | Only Qwen2.5 72B Instruct measured |
| SciCode | 54.2% | Not measured | Only Mistral Large 4 measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 4.5% | Only Qwen2.5 72B Instruct measured |
| FrontierCode | 39.4% | Not measured | Only Mistral Large 4 measured |
| WeirdML | Not measured | 16.0% | Only Qwen2.5 72B Instruct measured |
| ALE-Bench | 1213.2 | Not measured | Only Mistral Large 4 measured |
| LMArena Text: coding | 1478 | 1292 | Mistral Large 4 |
| LMArena Text: software and IT | 1467 | 1292 | Mistral Large 4 |
| LMArena WebDev: React | 1559 | Not measured | Only Mistral Large 4 measured |
| LMArena WebDev: HTML | 1487 | Not measured | Only Mistral Large 4 measured |
| Neither model has a result on the other 21 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| AA Intelligence Index | 38.4 | 7.7 | Mistral Large 4 |
| Epoch Capabilities Index | 150.0 | 129.0 | Mistral Large 4 |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index13 boards, none shared
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| LiveBench research | 28.1% | Not measured | Only Mistral Large 4 measured |
| LiveBench browser | 8.0% | Not measured | Only Mistral Large 4 measured |
| LiveBench long agentic | 11.8% | Not measured | Only Mistral Large 4 measured |
| Finance Agent (Vals AI) | 54.7% | Not measured | Only Mistral Large 4 measured |
| Tax Agent Bench (Vals AI) | 30.4% | Not measured | Only Mistral Large 4 measured |
| Legal Research Bench (Vals AI) | 31.7% | Not measured | Only Mistral Large 4 measured |
| Harvey's Legal Agent Benchmark (Vals AI) | 17.5% | Not measured | Only Mistral Large 4 measured |
| Excel Modeling Benchmark (Vals AI) | 56.2% | Not measured | Only Mistral Large 4 measured |
| BioMysteryBench (Vals AI) | 67.0% | Not measured | Only Mistral Large 4 measured |
| τ²-Bench Telecom | Not measured | 34.5% | Only Qwen2.5 72B Instruct measured |
| METR time horizons | Not measured | 35.8% | Only Qwen2.5 72B Instruct measured |
| BALROG | Not measured | 16.2% | Only Qwen2.5 72B Instruct measured |
| TheAgentCompany | Not measured | 5.7% | Only Qwen2.5 72B Instruct measured |
| Neither model has a result on the other 28 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| LiveBench reasoning | 83.9% | Not measured | Only Mistral Large 4 measured |
| MysteryMechanism (Vals AI) | 12.2% | Not measured | Only Mistral Large 4 measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 49.1% | Only Qwen2.5 72B Instruct measured |
| Humanity's Last Exam (Artificial Analysis run) | 35.0% | 3.6% | Mistral Large 4 |
| GPQA Diamond (Epoch AI run) | 88.1% | 49.1% | Mistral Large 4 |
| Chess puzzles | 33.0% | Not measured | Only Mistral Large 4 measured |
| Mystery game puzzles | 17.0% | Not measured | Only Mistral Large 4 measured |
| BIG-Bench Hard | Not measured | 79.8% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: hard prompts | 1444 | 1271 | Mistral Large 4 |
| LMArena Text: expert | 1444 | 1245 | Mistral Large 4 |
| LMArena Text: hard prompts in English | 1442 | 1279 | Mistral Large 4 |
| LMArena Vision: diagrams | 1288 | Not measured | Only Mistral Large 4 measured |
| Neither model has a result on the other 10 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index30
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| AA Math Index | Not measured | 14.0 | Only Qwen2.5 72B Instruct measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 14.0% | Only Qwen2.5 72B Instruct measured |
| OTIS Mock AIME | 98.3% | 8.1% | Mistral Large 4 |
| FrontierMath, tiers 1 to 3 | 58.6% | Not measured | Only Mistral Large 4 measured |
| FrontierMath, tier 4 | 26.8% | Not measured | Only Mistral Large 4 measured |
| ProofBench | 10.0% | Not measured | Only Mistral Large 4 measured |
| MATH level 5 | Not measured | 63.2% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: math | 1469 | 1283 | Mistral Large 4 |
| LMArena Text: mathematical work | 1462 | 1290 | Mistral Large 4 |
| Neither model has a result on the other 4 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| MedCode (Vals AI) | 40.7% | Not measured | Only Mistral Large 4 measured |
| MMLU Pro (Artificial Analysis run) | Not measured | 72.0% | Only Qwen2.5 72B Instruct measured |
| SimpleQA Verified | 20.0% | Not measured | Only Mistral Large 4 measured |
| MMLU | Not measured | 85.3% | Only Qwen2.5 72B Instruct measured |
| AI2 Reasoning Challenge | Not measured | 94.5% | Only Qwen2.5 72B Instruct measured |
| LMArena Text: science | 1448 | 1275 | Mistral Large 4 |
| LMArena Text: medicine and healthcare | 1433 | 1258 | Mistral Large 4 |
| LMArena Text: legal and government | 1425 | 1288 | Mistral Large 4 |
| LMArena Text: business and finance | 1439 | 1253 | Mistral Large 4 |
| Neither model has a result on the other 9 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| Design Arena: websites | 1258 | Not measured | Only Mistral Large 4 measured |
| Design Arena: UI components | 1275 | Not measured | Only Mistral Large 4 measured |
| Design Arena: game dev | 1226 | Not measured | Only Mistral Large 4 measured |
| Design Arena: data viz | 1274 | Not measured | Only Mistral Large 4 measured |
| Design Arena: 3D | 1253 | Not measured | Only Mistral Large 4 measured |
| Design Arena: all code categories | 1259 | Not measured | Only Mistral Large 4 measured |
| Design Arena: SVG | 1233 | Not measured | Only Mistral Large 4 measured |
| Design Arena: ASCII art | 1202 | Not measured | Only Mistral Large 4 measured |
| LiveBench design | 65.6% | Not measured | Only Mistral Large 4 measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index160
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | Leader |
|---|---|---|---|
| LiveBench data analysis | 76.5% | Not measured | Only Mistral Large 4 measured |
| MedScribe (Vals AI) | 80.4% | Not measured | Only Mistral Large 4 measured |
| IFBench | Not measured | 36.9% | Only Qwen2.5 72B Instruct measured |
| Long context reasoning (AA-LCR) | 81.3% | Not measured | Only Mistral Large 4 measured |
| DTBench | 88.2% | 62.9% | Mistral Large 4 |
| LMCA | 35.8% | 13.4% | Mistral Large 4 |
| Surface Evolver Bench | 28.1% | Not measured | Only Mistral Large 4 measured |
| Furniture assembly | 20.8% | Not measured | Only Mistral Large 4 measured |
| HellaSwag | Not measured | 84.8% | Only Qwen2.5 72B Instruct measured |
| WinoGrande | Not measured | 82.3% | Only Qwen2.5 72B Instruct measured |
| PIQA | Not measured | 82.6% | Only Qwen2.5 72B Instruct measured |
| LMArena Text | 1431 | 1269 | Mistral Large 4 |
| LMArena Text: instruction following | 1422 | 1254 | Mistral Large 4 |
| LMArena Text: creative writing | 1363 | 1221 | Mistral Large 4 |
| LMArena Vision | 1272 | Not measured | Only Mistral Large 4 measured |
| LMArena Text: English | 1433 | 1283 | Mistral Large 4 |
| LMArena Text: non-English | 1422 | 1252 | Mistral Large 4 |
| LMArena Text: Chinese | 1492 | 1273 | Mistral Large 4 |
| LMArena Text: German | Not measured | 1234 | Only Qwen2.5 72B Instruct measured |
| LMArena Text: Japanese | 1415 | 1180 | Mistral Large 4 |
| LMArena Text: Korean | Not measured | 1188 | Only Qwen2.5 72B Instruct measured |
| LMArena Text: French | 1492 | 1280 | Mistral Large 4 |
| LMArena Text: Spanish | 1455 | 1255 | Mistral Large 4 |
| LMArena Text: Russian | 1417 | 1264 | Mistral Large 4 |
| LMArena Text: multi-turn | 1435 | 1272 | Mistral Large 4 |
| LMArena Text: longer queries | 1428 | 1281 | Mistral Large 4 |
| LMArena Text: writing and language | 1386 | 1247 | Mistral Large 4 |
| LMArena Text: entertainment and media | 1378 | 1211 | Mistral Large 4 |
| LMArena Vision: English | 1283 | Not measured | Only Mistral Large 4 measured |
| LMArena Vision: reading text | 1284 | Not measured | Only Mistral Large 4 measured |
| Neither model has a result on the other 16 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Mistral Large 4 | Qwen2.5 72B Instruct | 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
Qwen2.5 72B Instruct costs less
Repository review
50K input + 3K output tokens
Qwen2.5 72B Instruct costs less
Agent loop
200K input + 10K output tokens
Qwen2.5 72B Instruct 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 Large 4 | Qwen2.5 72B Instruct |
|---|---|---|
| Maker | Mistral AI | Qwen |
| Weights | No public weights listed | Open weights |
| Released | October 6, 2026 | September 19, 2024 |
| Context window | 1.05M | 33K |
| Input $/Mtok | $1.36 | $0.36 |
| Output $/Mtok | $4.18 | $0.40 |
| Coding index | 56.4 | Not benchmarked |
| Independent measurements behind it | 5 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Mistral Large 4 better than Qwen2.5 72B Instruct for coding?
Only Mistral Large 4 has a coding score, 56.4 on the Vibe Coding Index; Qwen2.5 72B Instruct has none, so the two cannot be ranked on coding. Pick Qwen2.5 72B Instruct for output tokens 10.4x cheaper ($0.40 vs $4.18 per million).
Which is cheaper, Mistral Large 4 or Qwen2.5 72B Instruct?
Mistral Large 4 costs $1.36 per million input tokens and $4.18 per million output tokens; Qwen2.5 72B Instruct costs $0.36 and $0.40. Cost per solved task is known for neither of them.
Which has the bigger context window, Mistral Large 4 or Qwen2.5 72B Instruct?
Mistral Large 4, with 1.05M tokens against 33K.
How many benchmarks is the Mistral Large 4 vs Qwen2.5 72B Instruct comparison based on?
32 benchmarks have a result for both models as of October 21, 2026. Mistral Large 4's coding index rests on 5 independent benchmarks and Qwen2.5 72B Instruct's on 0. Where only one model has a result, the page says so instead of naming a winner.