Model comparison
Mixtral 8x7B vs Qwen2.5 72B Instruct
Qwen2.5 72B Instruct is the stronger model overall, scoring 31.5 to 26.8 on the Noometry Index. Mixtral 8x7B costs 3.5× less per token, which makes it the better buy when Qwen2.5 72B Instruct's lead doesn't matter for your workload.
Last verified . 33 shared benchmarks.
Mistral AI
Alibaba (Qwen)Summary
- They share 33 benchmarks with published results for both. Mixtral 8x7B scores higher in 0 categories and Qwen2.5 72B Instruct in 8 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Qwen2.5 72B Instruct leads 26.9 to 11.0.
- The biggest single-benchmark swing is MATH Level 5: 10% for Mixtral 8x7B and 63.2% for Qwen2.5 72B Instruct.
- Mixtral 8x7B is cheaper at $0.70 / $0.70 per million input/output tokens, against $1.40 / $5.60 for Qwen2.5 72B Instruct.
- Qwen2.5 72B Instruct accepts more context: 131K tokens versus 32K.
Side by side
| Mixtral 8x7B | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | Mistral AI | Alibaba (Qwen) |
| Noometry Index | 26.8 | 31.5 |
| Released | 2023-12-11 | 2024-09 |
| Weights | Open | Open |
| Context window | 32K | 131K |
| Max output | 32K | 8K |
| Input $ / M tokens | $0.70 | $1.40 |
| Output $ / M tokens | $0.70 | $5.60 |
| Results tracked | 38 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Mixtral 8x7B: 32.2 (#265), Qwen2.5 72B Instruct: 32.8 (#256)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Coding | 1127 | 1293 |
| WeirdML | — | 16% |
| BigCodeBench Instruct | — | 45.8% |
| BigCodeBench Complete | — | 55.9% |
| HumanEval+ | 39.6% | — |
| MBPP+ | 49.7% | — |
Agentic & Tool Use Not comparable
Mixtral 8x7B: —, Qwen2.5 72B Instruct: 22.0 (#124)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| METR Time Horizons | — | 35.8% |
Reasoning Qwen2.5 72B Instruct leads
Mixtral 8x7B: 17.4 (#283), Qwen2.5 72B Instruct: 21.4 (#197)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1115 | 1271 |
| DTBench | 49.6% | 62.9% |
| Epoch Capabilities Index | 118.38 | 129.01 |
| ForecastBench | 56.3 | 57.5 |
| HellaSwag | 86.7% | 84.8% |
| PIQA | 83.6% | 82.6% |
| WinoGrande | 77.2% | 82.3% |
| LMCA | — | 13.4% |
| Adversarial NLI | 55.2% | — |
| BIG-Bench Hard | — | 79.8% |
Math Too close to call
Mixtral 8x7B: 17.2 (#284), Qwen2.5 72B Instruct: 18.1 (#281)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| Omni-MATH | 10.5% | 33% |
| LMArena Math | 1147 | 1283 |
| MATH Level 5 | 10% | 63.2% |
| OTIS Mock AIME 2024-2025 | — | 8.1% |
| GSM8K | 74.4% | — |
Knowledge Qwen2.5 72B Instruct leads
Mixtral 8x7B: 11.0 (#296), Qwen2.5 72B Instruct: 26.9 (#247)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | 30.6% | 49.1% |
| MMLU-Pro | 33.5% | 63.1% |
| GPQA (HELM) | 29.6% | 42.6% |
| LMArena Expert | 1088 | 1245 |
| ARC (AI2) Challenge | 87.3% | 94.5% |
| MMLU | 70.6% | 85.3% |
| TriviaQA | 82.2% | 71.9% |
| Confabulations | — | 19.1% |
| OpenBookQA | 85.8% | — |
Multilingual Qwen2.5 72B Instruct leads
Mixtral 8x7B: 29.8 (#261), Qwen2.5 72B Instruct: 41.2 (#208)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1077 | 1252 |
| LMArena Chinese | 1055 | 1272 |
| LMArena French | 1165 | 1279 |
| LMArena German | 1114 | 1234 |
| LMArena Japanese | 931 | 1180 |
| LMArena Korean | 968 | 1188 |
| LMArena Russian | 1090 | 1264 |
| LMArena Spanish | 1110 | 1255 |
Instruction Following Qwen2.5 72B Instruct leads
Mixtral 8x7B: 51.2 (#292), Qwen2.5 72B Instruct: 65.7 (#216)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| IFEval | 57.5% | 80.6% |
| LMArena Instruction Following | 1110 | 1254 |
Long Context Qwen2.5 72B Instruct leads
Mixtral 8x7B: 33.4 (#255), Qwen2.5 72B Instruct: 38.9 (#183)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1103 | 1282 |
Writing & Preference Qwen2.5 72B Instruct leads
Mixtral 8x7B: 34.4 (#265), Qwen2.5 72B Instruct: 46.9 (#210)
| Benchmark | Mixtral 8x7B | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1132 | 1269 |
| LMArena Creative Writing | 1109 | 1221 |
| WildBench | 67.3% | 80.2% |
| LMArena Multi-Turn | 1115 | 1272 |
Frequently asked questions
Is Mixtral 8x7B better than Qwen2.5 72B Instruct?
Qwen2.5 72B Instruct is the stronger model overall, scoring 31.5 to 26.8 on the Noometry Index. Mixtral 8x7B costs 3.5× less per token, which makes it the better buy when Qwen2.5 72B Instruct's lead doesn't matter for your workload.
Which is cheaper, Mixtral 8x7B or Qwen2.5 72B Instruct?
Mixtral 8x7B is cheaper. It lists at $0.70 per million input tokens and $0.70 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.
Is Mixtral 8x7B or Qwen2.5 72B Instruct better for coding?
They score almost the same on coding (32.2 vs 32.8); test both on your own repository before choosing.
Which has the bigger context window?
Qwen2.5 72B Instruct does, with 131K tokens against 32K.
How many benchmarks do Mixtral 8x7B and Qwen2.5 72B Instruct share?
33 benchmarks have published results for both models. Mixtral 8x7B has 38 scored results on Noometry and Qwen2.5 72B Instruct has 43.