Model comparison
Llama 3.1-70B vs Mixtral 8x7B
Llama 3.1-70B is the stronger model overall, scoring 29.4 to 26.8 on the Noometry Index.
Last verified . 27 shared benchmarks.
Meta
Mistral AISummary
- They share 27 benchmarks with published results for both. Llama 3.1-70B scores higher in 6 categories and Mixtral 8x7B in 2 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in instruction following, where Llama 3.1-70B leads 65.5 to 51.2.
- The biggest single-benchmark swing is MMLU-Pro: 65.3% for Llama 3.1-70B and 33.5% for Mixtral 8x7B.
- Llama 3.1-70B is cheaper at $0.40 / $0.40 per million input/output tokens, against $0.70 / $0.70 for Mixtral 8x7B.
- Llama 3.1-70B accepts more context: 128K tokens versus 32K.
Side by side
| Llama 3.1-70B | Mixtral 8x7B | |
|---|---|---|
| Provider | Meta | Mistral AI |
| Noometry Index | 29.4 | 26.8 |
| Released | 2024-07-23 | 2023-12-11 |
| Weights | Open | Open |
| Context window | 128K | 32K |
| Max output | 4K | 32K |
| Input $ / M tokens | $0.40 | $0.70 |
| Output $ / M tokens | $0.40 | $0.70 |
| Results tracked | 35 | 38 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Mixtral 8x7B leads
Llama 3.1-70B: 29.8 (#292), Mixtral 8x7B: 32.2 (#265)
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| LMArena Coding | 1260 | 1127 |
| WeirdML | 9% | — |
| BigCodeBench Instruct | 46.1% | — |
| BigCodeBench Complete | 54.8% | — |
| HumanEval+ | — | 39.6% |
| MBPP+ | — | 49.7% |
Agentic & Tool Use Not comparable
Llama 3.1-70B: 24.9 (#104), Mixtral 8x7B: —
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| TheAgentCompany | 6.9% | — |
| BALROG | 27.9% | — |
Reasoning Llama 3.1-70B leads
Llama 3.1-70B: 20.8 (#215), Mixtral 8x7B: 17.4 (#283)
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| LMArena Hard Prompts | 1241 | 1115 |
| DTBench | 60% | 49.6% |
| Epoch Capabilities Index | 125.91 | 118.38 |
| LMCA | 14.8% | — |
| Adversarial NLI | — | 55.2% |
| ForecastBench | — | 56.3 |
| HellaSwag | — | 86.7% |
| PIQA | — | 83.6% |
| WinoGrande | — | 77.2% |
Math Mixtral 8x7B leads
Llama 3.1-70B: 12.7 (#298), Mixtral 8x7B: 17.2 (#284)
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| Omni-MATH | 21% | 10.5% |
| LMArena Math | 1252 | 1147 |
| MATH Level 5 | 36.7% | 10% |
| OTIS Mock AIME 2024-2025 | 3.6% | — |
| GSM8K | — | 74.4% |
Knowledge Llama 3.1-70B leads
Llama 3.1-70B: 24.1 (#264), Mixtral 8x7B: 11.0 (#296)
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| GPQA Diamond | 44.2% | 30.6% |
| MMLU-Pro | 65.3% | 33.5% |
| GPQA (HELM) | 42.6% | 29.6% |
| LMArena Expert | 1209 | 1088 |
| MMLU | 80.1% | 70.6% |
| ARC (AI2) Challenge | — | 87.3% |
| OpenBookQA | — | 85.8% |
| TriviaQA | — | 82.2% |
Multilingual Llama 3.1-70B leads
Llama 3.1-70B: 38.9 (#220), Mixtral 8x7B: 29.8 (#261)
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| LMArena Non-English | 1219 | 1077 |
| LMArena Chinese | 1215 | 1055 |
| LMArena French | 1260 | 1165 |
| LMArena German | 1222 | 1114 |
| LMArena Japanese | 1132 | 931 |
| LMArena Korean | 1140 | 968 |
| LMArena Russian | 1234 | 1090 |
| LMArena Spanish | 1253 | 1110 |
Instruction Following Llama 3.1-70B leads
Llama 3.1-70B: 65.5 (#218), Mixtral 8x7B: 51.2 (#292)
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| IFEval | 82.1% | 57.5% |
| LMArena Instruction Following | 1231 | 1110 |
Long Context Llama 3.1-70B leads
Llama 3.1-70B: 37.6 (#208), Mixtral 8x7B: 33.4 (#255)
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| LMArena Longer Query | 1241 | 1103 |
Writing & Preference Llama 3.1-70B leads
Llama 3.1-70B: 35.6 (#262), Mixtral 8x7B: 34.4 (#265)
| Benchmark | Llama 3.1-70B | Mixtral 8x7B |
|---|---|---|
| LMArena Text | 1261 | 1132 |
| LMArena Creative Writing | 1232 | 1109 |
| WildBench | 75.8% | 67.3% |
| LMArena Multi-Turn | 1256 | 1115 |
| EQ-Bench Creative Writing | 784 | — |
Frequently asked questions
Is Llama 3.1-70B better than Mixtral 8x7B?
Llama 3.1-70B is the stronger model overall, scoring 29.4 to 26.8 on the Noometry Index.
Which is cheaper, Llama 3.1-70B or Mixtral 8x7B?
Llama 3.1-70B is cheaper. It lists at $0.40 per million input tokens and $0.40 per million output tokens; Mixtral 8x7B lists at $0.70 and $0.70.
Is Llama 3.1-70B or Mixtral 8x7B better for coding?
Mixtral 8x7B scores higher on coding benchmarks: 32.2 versus 29.8 in the Noometry coding category.
Which has the bigger context window?
Llama 3.1-70B does, with 128K tokens against 32K.
How many benchmarks do Llama 3.1-70B and Mixtral 8x7B share?
27 benchmarks have published results for both models. Llama 3.1-70B has 35 scored results on Noometry and Mixtral 8x7B has 38.