Model comparison
DeepSeek-V3 vs Mixtral 8x7B
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 27.1 on the Noometry Index.
Last verified . 35 shared benchmarks.
DeepSeek
Mistral AISummary
- They share 35 benchmarks with published results for both. DeepSeek-V3 scores higher in 8 categories and Mixtral 8x7B in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3 leads 37.5 to 11.0.
- The biggest single-benchmark swing is MATH Level 5: 75.5% for DeepSeek-V3 and 10% for Mixtral 8x7B.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $0.70 / $0.70 for Mixtral 8x7B.
- DeepSeek-V3 accepts more context: 164K tokens versus 32K.
Side by side
| DeepSeek-V3 | Mixtral 8x7B | |
|---|---|---|
| Provider | DeepSeek | Mistral AI |
| Noometry Index | 39.5 | 27.1 |
| Released | 2024-12-26 | 2023-12-11 |
| Weights | Open | Open |
| Context window | 164K | 32K |
| Max output | 164K | 32K |
| Input $ / M tokens | $0.24 | $0.70 |
| Output $ / M tokens | $0.90 | $0.70 |
| Results tracked | 60 | 38 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
DeepSeek-V3: 42.3 (#106), Mixtral 8x7B: 32.8 (#269)
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| LMArena Coding | 1368 | 1126 |
| HumanEval+ | 86.6% | 39.6% |
| MBPP+ | 73% | 49.7% |
| Aider Polyglot | 55.1% | — |
| SciCode | 35.8% | — |
| WeirdML | 36.1% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Mixtral 8x7B: —
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| METR Time Horizons | 49.6% | — |
Reasoning DeepSeek-V3 leads
DeepSeek-V3: 20.5 (#236), Mixtral 8x7B: 18.2 (#285)
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| LMArena Hard Prompts | 1365 | 1115 |
| DTBench | 64.8% | 49.6% |
| Epoch Capabilities Index | 135.94 | 118.47 |
| ForecastBench | 59.1 | 56.3 |
| HellaSwag | 88.9% | 86.7% |
| PIQA | 84.7% | 83.6% |
| WinoGrande | 85.2% | 77.2% |
| SimpleBench | 27.2% | — |
| Kagi LLM Benchmark | 52.3% | — |
| CritPt | 0% | — |
| LiveBench Reasoning | 65.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| LMCA | 15.5% | — |
| Adversarial NLI | — | 55.2% |
| BIG-Bench Hard | 87.5% | — |
| LiveBench | 66.9% | — |
Math DeepSeek-V3 leads
DeepSeek-V3: 32.1 (#219), Mixtral 8x7B: 18.8 (#289)
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| Omni-MATH | 40.3% | 10.5% |
| LMArena Math | 1373 | 1147 |
| MATH Level 5 | 75.5% | 10% |
| OTIS Mock AIME 2024-2025 | 37.8% | — |
| LiveBench Math | 73.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
| GSM8K | — | 74.4% |
Knowledge DeepSeek-V3 leads
DeepSeek-V3: 37.5 (#155), Mixtral 8x7B: 11.0 (#301)
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| GPQA Diamond | 67.6% | 30.6% |
| MMLU-Pro | 72.3% | 33.5% |
| GPQA (HELM) | 53.8% | 29.6% |
| LMArena Expert | 1351 | 1088 |
| ARC (AI2) Challenge | 95.3% | 87.3% |
| MMLU | 87.2% | 70.6% |
| TriviaQA | 82.9% | 82.2% |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| OpenBookQA | — | 85.8% |
Multilingual DeepSeek-V3 leads
DeepSeek-V3: 48.5 (#143), Mixtral 8x7B: 29.6 (#266)
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| LMArena Non-English | 1358 | 1077 |
| LMArena Chinese | 1391 | 1055 |
| LMArena French | 1385 | 1166 |
| LMArena German | 1374 | 1114 |
| LMArena Japanese | 1333 | 931 |
| LMArena Korean | 1319 | 968 |
| LMArena Russian | 1373 | 1090 |
| LMArena Spanish | 1358 | 1111 |
Instruction Following DeepSeek-V3 leads
DeepSeek-V3: 72.8 (#130), Mixtral 8x7B: 51.0 (#297)
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| IFEval | 83.2% | 57.5% |
| LMArena Instruction Following | 1345 | 1109 |
| LiveBench Instruction Following | 81.5% | — |
Long Context Too close to call
DeepSeek-V3: 34.0 (#253), Mixtral 8x7B: 33.4 (#260)
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| LMArena Longer Query | 1352 | 1103 |
| Fiction.LiveBench | 50% | — |
Writing & Preference DeepSeek-V3 leads
DeepSeek-V3: 57.4 (#130), Mixtral 8x7B: 34.2 (#270)
| Benchmark | DeepSeek-V3 | Mixtral 8x7B |
|---|---|---|
| LMArena Text | 1375 | 1132 |
| LMArena Creative Writing | 1364 | 1109 |
| WildBench | 83% | 67.3% |
| LMArena Multi-Turn | 1389 | 1115 |
| Short-Story Creative Writing | 77% | — |
| EQ-Bench Creative Writing | 1472 | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than Mixtral 8x7B?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 27.1 on the Noometry Index.
Which is cheaper, DeepSeek-V3 or Mixtral 8x7B?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Mixtral 8x7B lists at $0.70 and $0.70.
Is DeepSeek-V3 or Mixtral 8x7B better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 versus 32.8 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3 does, with 164K tokens against 32K.
How many benchmarks do DeepSeek-V3 and Mixtral 8x7B share?
35 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Mixtral 8x7B has 38.