Model comparison
Mixtral 8x7B vs Nemotron 4 340b Instruct
Nemotron 4 340b Instruct is the stronger model overall, scoring 35.3 to 26.8 on the Noometry Index.
Last verified . 17 shared benchmarks.
Mistral AI
NVIDIASummary
- They share 17 benchmarks with published results for both. Mixtral 8x7B scores higher in 0 categories and Nemotron 4 340b Instruct in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Nemotron 4 340b Instruct leads 31.4 to 11.0.
Side by side
| Mixtral 8x7B | Nemotron 4 340b Instruct | |
|---|---|---|
| Provider | Mistral AI | NVIDIA |
| Noometry Index | 26.8 | 35.3 |
| Released | 2023-12-11 | 2024-06-14 |
| Weights | Open | Open |
| Context window | 32K | — |
| Max output | 32K | — |
| Input $ / M tokens | $0.70 | — |
| Output $ / M tokens | $0.70 | — |
| Results tracked | 38 | 17 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Nemotron 4 340b Instruct leads
Mixtral 8x7B: 32.2 (#265), Nemotron 4 340b Instruct: 34.6 (#223)
| Benchmark | Mixtral 8x7B | Nemotron 4 340b Instruct |
|---|---|---|
| LMArena Coding | 1127 | 1209 |
| HumanEval+ | 39.6% | — |
| MBPP+ | 49.7% | — |
Reasoning Nemotron 4 340b Instruct leads
Mixtral 8x7B: 17.4 (#283), Nemotron 4 340b Instruct: 22.4 (#180)
| Benchmark | Mixtral 8x7B | Nemotron 4 340b Instruct |
|---|---|---|
| LMArena Hard Prompts | 1115 | 1202 |
| DTBench | 49.6% | — |
| Adversarial NLI | 55.2% | — |
| Epoch Capabilities Index | 118.38 | — |
| ForecastBench | 56.3 | — |
| HellaSwag | 86.7% | — |
| PIQA | 83.6% | — |
| WinoGrande | 77.2% | — |
Math Nemotron 4 340b Instruct leads
Mixtral 8x7B: 17.2 (#284), Nemotron 4 340b Instruct: 31.7 (#193)
| Benchmark | Mixtral 8x7B | Nemotron 4 340b Instruct |
|---|---|---|
| LMArena Math | 1147 | 1216 |
| Omni-MATH | 10.5% | — |
| MATH Level 5 | 10% | — |
| GSM8K | 74.4% | — |
Knowledge Nemotron 4 340b Instruct leads
Mixtral 8x7B: 11.0 (#296), Nemotron 4 340b Instruct: 31.4 (#212)
| Benchmark | Mixtral 8x7B | Nemotron 4 340b Instruct |
|---|---|---|
| LMArena Expert | 1088 | 1171 |
| GPQA Diamond | 30.6% | — |
| MMLU-Pro | 33.5% | — |
| GPQA (HELM) | 29.6% | — |
| ARC (AI2) Challenge | 87.3% | — |
| MMLU | 70.6% | — |
| OpenBookQA | 85.8% | — |
| TriviaQA | 82.2% | — |
Multilingual Nemotron 4 340b Instruct leads
Mixtral 8x7B: 29.8 (#261), Nemotron 4 340b Instruct: 38.0 (#228)
| Benchmark | Mixtral 8x7B | Nemotron 4 340b Instruct |
|---|---|---|
| LMArena Non-English | 1077 | 1206 |
| LMArena Chinese | 1055 | 1214 |
| LMArena French | 1165 | 1196 |
| LMArena German | 1114 | 1192 |
| LMArena Japanese | 931 | 1128 |
| LMArena Korean | 968 | 1148 |
| LMArena Russian | 1090 | 1219 |
| LMArena Spanish | 1110 | 1195 |
Instruction Following Nemotron 4 340b Instruct leads
Mixtral 8x7B: 51.2 (#292), Nemotron 4 340b Instruct: 63.3 (#227)
| Benchmark | Mixtral 8x7B | Nemotron 4 340b Instruct |
|---|---|---|
| LMArena Instruction Following | 1110 | 1203 |
| IFEval | 57.5% | — |
Long Context Nemotron 4 340b Instruct leads
Mixtral 8x7B: 33.4 (#255), Nemotron 4 340b Instruct: 37.1 (#216)
| Benchmark | Mixtral 8x7B | Nemotron 4 340b Instruct |
|---|---|---|
| LMArena Longer Query | 1103 | 1225 |
Writing & Preference Nemotron 4 340b Instruct leads
Mixtral 8x7B: 34.4 (#265), Nemotron 4 340b Instruct: 42.9 (#229)
| Benchmark | Mixtral 8x7B | Nemotron 4 340b Instruct |
|---|---|---|
| LMArena Text | 1132 | 1225 |
| LMArena Creative Writing | 1109 | 1203 |
| LMArena Multi-Turn | 1115 | 1214 |
| WildBench | 67.3% | — |
Frequently asked questions
Is Mixtral 8x7B better than Nemotron 4 340b Instruct?
Nemotron 4 340b Instruct is the stronger model overall, scoring 35.3 to 26.8 on the Noometry Index.
Is Mixtral 8x7B or Nemotron 4 340b Instruct better for coding?
Nemotron 4 340b Instruct scores higher on coding benchmarks: 34.6 versus 32.2 in the Noometry coding category.
How many benchmarks do Mixtral 8x7B and Nemotron 4 340b Instruct share?
17 benchmarks have published results for both models. Mixtral 8x7B has 38 scored results on Noometry and Nemotron 4 340b Instruct has 17.