Model comparison
Llama 3-8B vs Mixtral 8x22B
Mixtral 8x22B is the stronger model overall, scoring 27.1 to 25.5 on the Noometry Index.
Last verified . 27 shared benchmarks.
Meta
Mistral AISummary
- They share 27 benchmarks with published results for both. Llama 3-8B scores higher in 3 categories and Mixtral 8x22B in 5 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where Mixtral 8x22B leads 22.9 to 8.8.
- The biggest single-benchmark swing is MATH Level 5: 6.1% for Llama 3-8B and 24.2% for Mixtral 8x22B.
Side by side
| Llama 3-8B | Mixtral 8x22B | |
|---|---|---|
| Provider | Meta | Mistral AI |
| Noometry Index | 25.5 | 27.1 |
| Released | 2024-04-18 | 2024-04-17 |
| Weights | Open | Open |
| Context window | — | 64K |
| Max output | — | 64K |
| Input $ / M tokens | — | $2 |
| Output $ / M tokens | — | $6 |
| Results tracked | 34 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Llama 3-8B leads
Llama 3-8B: 31.0 (#289), Mixtral 8x22B: 24.2 (#329)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| BigCodeBench Instruct | 31.9% | 40.6% |
| LMArena Coding | 1152 | 1166 |
| BigCodeBench Complete | 36.9% | 50.2% |
| HumanEval+ | 56.7% | 72% |
| MBPP+ | 54.8% | 64.3% |
| WeirdML | — | 3.2% |
Agentic & Tool Use Not comparable
Llama 3-8B: —, Mixtral 8x22B: 23.1 (#127)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| Cybench | — | 7.5% |
Reasoning Mixtral 8x22B leads
Llama 3-8B: 14.3 (#326), Mixtral 8x22B: 19.9 (#248)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| LMArena Hard Prompts | 1133 | 1150 |
| DTBench | 43.9% | 55.1% |
| Epoch Capabilities Index | 116.45 | 122.03 |
| ForecastBench | 58.6 | 56.3 |
| Chess Puzzles | 0% | — |
| Adversarial NLI | 57.3% | — |
| WinoGrande | 75.7% | — |
Math Mixtral 8x22B leads
Llama 3-8B: 8.8 (#323), Mixtral 8x22B: 22.9 (#275)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| LMArena Math | 1151 | 1184 |
| MATH Level 5 | 6.1% | 24.2% |
| OTIS Mock AIME 2024-2025 | 1.9% | — |
| Omni-MATH | — | 16.3% |
Knowledge Mixtral 8x22B leads
Llama 3-8B: 7.8 (#308), Mixtral 8x22B: 15.1 (#293)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| GPQA Diamond | 26.1% | 34.1% |
| LMArena Expert | 1113 | 1113 |
| MMLU | 68.8% | 77.8% |
| MMLU-Pro | — | 46% |
| GPQA (HELM) | — | 33.4% |
| ARC (AI2) Challenge | 82.8% | — |
| OpenBookQA | 82.6% | — |
| TriviaQA | 67.7% | — |
Multilingual Mixtral 8x22B leads
Llama 3-8B: 30.8 (#261), Mixtral 8x22B: 32.8 (#255)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| LMArena Non-English | 1098 | 1128 |
| LMArena Chinese | 1076 | 1116 |
| LMArena French | 1159 | 1166 |
| LMArena German | 1104 | 1141 |
| LMArena Japanese | 967 | 1037 |
| LMArena Korean | 1004 | 1057 |
| LMArena Russian | 1109 | 1158 |
| LMArena Spanish | 1173 | 1151 |
Instruction Following Too close to call
Llama 3-8B: 58.4 (#260), Mixtral 8x22B: 57.7 (#266)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| LMArena Instruction Following | 1127 | 1147 |
| IFEval | — | 72.4% |
Long Context Too close to call
Llama 3-8B: 34.2 (#251), Mixtral 8x22B: 34.7 (#247)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| LMArena Longer Query | 1128 | 1144 |
Writing & Preference Too close to call
Llama 3-8B: 37.5 (#256), Mixtral 8x22B: 36.9 (#262)
| Benchmark | Llama 3-8B | Mixtral 8x22B |
|---|---|---|
| LMArena Text | 1166 | 1162 |
| LMArena Creative Writing | 1150 | 1141 |
| LMArena Multi-Turn | 1152 | 1130 |
| WildBench | — | 71.1% |
Frequently asked questions
Is Llama 3-8B better than Mixtral 8x22B?
Mixtral 8x22B is the stronger model overall, scoring 27.1 to 25.5 on the Noometry Index.
Is Llama 3-8B or Mixtral 8x22B better for coding?
Llama 3-8B scores higher on coding benchmarks: 31.0 versus 24.2 in the Noometry coding category.
How many benchmarks do Llama 3-8B and Mixtral 8x22B share?
27 benchmarks have published results for both models. Llama 3-8B has 34 scored results on Noometry and Mixtral 8x22B has 34.