Model comparison
DeepSeek-V2 (MoE-236B, May 2024) vs Llama 3.1-405B
Llama 3.1-405B has enough public results to be ranked (#288); DeepSeek-V2 (MoE-236B, May 2024) does not yet, so treat this comparison as directional.
Last verified . 8 shared benchmarks.
DeepSeek
MetaSummary
- They share 8 benchmarks with published results for both. DeepSeek-V2 (MoE-236B, May 2024) scores higher in 1 category and Llama 3.1-405B in 0 categories; one gap is clear of the uncertainty.
- The widest gap is in coding, where DeepSeek-V2 (MoE-236B, May 2024) leads 40.4 to 33.1.
Side by side
| DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 40.3 | 30.7 |
| Released | 2024-05-07 | 2024-07-23 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 10 | 42 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V2 (MoE-236B, May 2024) leads
DeepSeek-V2 (MoE-236B, May 2024): 40.4 (#139), Llama 3.1-405B: 33.1 (#262)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| WeirdML | — | 21.4% |
| BigCodeBench Instruct | 48.9% | — |
| LMArena Coding | — | 1291 |
| BigCodeBench Complete | 59.4% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.1-405B: 21.0 (#140)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| TheAgentCompany | — | 7.4% |
| Cybench | — | 7.5% |
Reasoning Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.1-405B: 16.8 (#300)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| BIG-Bench Hard | 78.8% | 82.9% |
| Epoch Capabilities Index | 124.77 | 128.75 |
| HellaSwag | 87.1% | 89.2% |
| PIQA | 83.9% | 85.9% |
| WinoGrande | 86.3% | 89.2% |
| SimpleBench | — | 23% |
| Kagi LLM Benchmark | — | 45% |
| LMArena Hard Prompts | — | 1269 |
| DTBench | — | 61.4% |
| ForecastBench | — | 59.9 |
Math Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.1-405B: 18.4 (#290)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | — | 9.7% |
| Omni-MATH | — | 24.9% |
| LMArena Math | — | 1281 |
| MATH Level 5 | — | 49.8% |
Knowledge Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.1-405B: 30.4 (#227)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| ARC (AI2) Challenge | 92.2% | 95.3% |
| MMLU | 78.4% | 84.5% |
| TriviaQA | 80% | 82.7% |
| GPQA Diamond | — | 50.9% |
| MMLU-Pro | — | 72.3% |
| Confabulations | — | 17.6% |
| GPQA (HELM) | — | 52.2% |
| LMArena Expert | — | 1243 |
Multilingual Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.1-405B: 40.7 (#214)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| LMArena Non-English | — | 1248 |
| LMArena Chinese | — | 1242 |
| LMArena French | — | 1279 |
| LMArena German | — | 1252 |
| LMArena Japanese | — | 1208 |
| LMArena Korean | — | 1184 |
| LMArena Russian | — | 1265 |
| LMArena Spanish | — | 1260 |
Instruction Following Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.1-405B: 65.9 (#214)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| IFEval | — | 81.1% |
| LMArena Instruction Following | — | 1259 |
Long Context Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.1-405B: 38.4 (#197)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| LMArena Longer Query | — | 1266 |
Writing & Preference Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.1-405B: 38.9 (#251)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.1-405B |
|---|---|---|
| LMArena Text | — | 1284 |
| LMArena Creative Writing | — | 1262 |
| EQ-Bench Creative Writing | — | 870 |
| WildBench | — | 78.3% |
| LMArena Multi-Turn | — | 1297 |
Frequently asked questions
Is DeepSeek-V2 (MoE-236B, May 2024) better than Llama 3.1-405B?
Llama 3.1-405B has enough public results to be ranked (#288); DeepSeek-V2 (MoE-236B, May 2024) does not yet, so treat this comparison as directional.
Is DeepSeek-V2 (MoE-236B, May 2024) or Llama 3.1-405B better for coding?
DeepSeek-V2 (MoE-236B, May 2024) scores higher on coding benchmarks: 40.4 versus 33.1 in the Noometry coding category.
How many benchmarks do DeepSeek-V2 (MoE-236B, May 2024) and Llama 3.1-405B share?
8 benchmarks have published results for both models. DeepSeek-V2 (MoE-236B, May 2024) has 10 scored results on Noometry and Llama 3.1-405B has 42.