Model comparison
Mistral Small vs Qwen2.5 72B Instruct
Mistral Small is the stronger model overall, scoring 33.4 to 31.9 on the Noometry Index.
Last verified . 25 shared benchmarks.
Mistral AI
Alibaba (Qwen)Summary
- They share 25 benchmarks with published results for both. Mistral Small scores higher in 7 categories and Qwen2.5 72B Instruct in 2 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where Mistral Small leads 28.1 to 22.1.
- The biggest single-benchmark swing is MATH Level 5: 46.8% for Mistral Small and 63.2% for Qwen2.5 72B Instruct.
- Mistral Small is cheaper at $0.15 / $0.60 per million input/output tokens, against $1.40 / $5.60 for Qwen2.5 72B Instruct.
- Mistral Small accepts more context: 262K tokens versus 131K.
Side by side
| Mistral Small | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | Mistral AI | Alibaba (Qwen) |
| Noometry Index | 33.4 | 31.9 |
| Released | 2024-02-26 | 2024-09 |
| Weights | Open | Open |
| Context window | 262K | 131K |
| Max output | 256K | 8K |
| Input $ / M tokens | $0.15 | $1.40 |
| Output $ / M tokens | $0.60 | $5.60 |
| Results tracked | 39 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Mistral Small: 34.0 (#247), Qwen2.5 72B Instruct: 33.2 (#260)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| BigCodeBench Instruct | 36.1% | 45.8% |
| LMArena Coding | 1362 | 1292 |
| BigCodeBench Complete | 46.6% | 55.9% |
| SciCode | 26.5% | — |
| WeirdML | — | 16% |
| LiveBench Coding | 36.2% | — |
| ALE-Bench | 497.62 | — |
Agentic & Tool Use Mistral Small leads
Mistral Small: 28.1 (#93), Qwen2.5 72B Instruct: 22.1 (#133)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| Berkeley Function Calling Leaderboard | 37.1% | — |
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| METR Time Horizons | — | 35.8% |
Reasoning Qwen2.5 72B Instruct leads
Mistral Small: 19.8 (#250), Qwen2.5 72B Instruct: 22.3 (#199)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1335 | 1271 |
| DTBench | 70.9% | 62.9% |
| LMCA | 20.6% | 13.4% |
| Kagi LLM Benchmark | 37.8% | — |
| CritPt | 0% | — |
| LiveBench Reasoning | 44.8% | — |
| LiveBench Data Analysis | 53.7% | — |
| BIG-Bench Hard | — | 79.8% |
| Epoch Capabilities Index | — | 129 |
| ForecastBench | — | 57.5 |
| HellaSwag | — | 84.8% |
| LiveBench | 44% | — |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math Qwen2.5 72B Instruct leads
Mistral Small: 16.4 (#293), Qwen2.5 72B Instruct: 19.3 (#287)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 5.8% | 8.1% |
| LMArena Math | 1341 | 1283 |
| MATH Level 5 | 46.8% | 63.2% |
| Omni-MATH | — | 33% |
| LiveBench Math | 39.9% | — |
Knowledge Mistral Small leads
Mistral Small: 31.0 (#222), Qwen2.5 72B Instruct: 27.0 (#253)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | 47.5% | 49.1% |
| LMArena Expert | 1291 | 1245 |
| MMLU | 68.7% | 85.3% |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| Vectara Hallucination Rate | 5.1% | — |
| GPQA (HELM) | — | 42.6% |
| ARC (AI2) Challenge | — | 94.5% |
| TriviaQA | — | 71.9% |
Multimodal Not comparable
Mistral Small: 33.5 (#96), Qwen2.5 72B Instruct: —
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Vision | 1142 | — |
Multilingual Mistral Small leads
Mistral Small: 45.5 (#169), Qwen2.5 72B Instruct: 41.0 (#213)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1315 | 1252 |
| LMArena Chinese | 1340 | 1272 |
| LMArena French | 1337 | 1280 |
| LMArena German | 1340 | 1234 |
| LMArena Japanese | 1275 | 1180 |
| LMArena Korean | 1259 | 1188 |
| LMArena Russian | 1324 | 1264 |
| LMArena Spanish | 1346 | 1256 |
Instruction Following Too close to call
Mistral Small: 66.4 (#209), Qwen2.5 72B Instruct: 65.5 (#221)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1310 | 1254 |
| LiveBench Instruction Following | 63.7% | — |
| IFEval | — | 80.6% |
Long Context Mistral Small leads
Mistral Small: 40.4 (#156), Qwen2.5 72B Instruct: 38.9 (#188)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1327 | 1282 |
Writing & Preference Mistral Small leads
Mistral Small: 52.5 (#171), Qwen2.5 72B Instruct: 46.7 (#215)
| Benchmark | Mistral Small | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1338 | 1269 |
| LMArena Creative Writing | 1305 | 1221 |
| LMArena Multi-Turn | 1344 | 1272 |
| WildBench | — | 80.2% |
| LiveBench Language | 30.5% | — |
Frequently asked questions
Is Mistral Small better than Qwen2.5 72B Instruct?
Mistral Small is the stronger model overall, scoring 33.4 to 31.9 on the Noometry Index.
Which is cheaper, Mistral Small or Qwen2.5 72B Instruct?
Mistral Small is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.
Is Mistral Small or Qwen2.5 72B Instruct better for coding?
They score almost the same on coding (34.0 vs 33.2); test both on your own repository before choosing.
Which has the bigger context window?
Mistral Small does, with 262K tokens against 131K.
How many benchmarks do Mistral Small and Qwen2.5 72B Instruct share?
25 benchmarks have published results for both models. Mistral Small has 39 scored results on Noometry and Qwen2.5 72B Instruct has 43.