Model comparison
DeepSeek-V2 (MoE-236B, May 2024) vs Llama 3.2 3B
Llama 3.2 3B has enough public results to be ranked (#321); DeepSeek-V2 (MoE-236B, May 2024) does not yet, so treat this comparison as directional.
Last verified . 2 shared benchmarks.
DeepSeek
MetaSummary
- They share 2 benchmarks with published results for both. DeepSeek-V2 (MoE-236B, May 2024) scores higher in 1 category and Llama 3.2 3B 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 27.6.
- The biggest single-benchmark swing is BigCodeBench Complete: 59.4% for DeepSeek-V2 (MoE-236B, May 2024) and 28.3% for Llama 3.2 3B.
Side by side
| DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 40.3 | 28.9 |
| Released | 2024-05-07 | 2024-09-24 |
| Weights | Open | Open |
| Context window | — | 131K |
| Max output | — | 118K |
| Input $ / M tokens | — | $0.05 |
| Output $ / M tokens | — | $0.33 |
| Results tracked | 10 | 18 |
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 3.2 3B: 27.6 (#319)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| BigCodeBench Instruct | 48.9% | 23.4% |
| BigCodeBench Complete | 59.4% | 28.3% |
| LMArena Coding | — | 1098 |
Agentic & Tool Use Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.2 3B: 20.1 (#143)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 21.9% |
| BALROG | — | 10.1% |
Reasoning Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.2 3B: 21.0 (#228)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| LMArena Hard Prompts | — | 1095 |
| BIG-Bench Hard | 78.8% | — |
| Epoch Capabilities Index | 124.77 | — |
| HellaSwag | 87.1% | — |
| PIQA | 83.9% | — |
| WinoGrande | 86.3% | — |
Math Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.2 3B: 32.4 (#214)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| LMArena Math | — | 1126 |
Knowledge Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.2 3B: 29.7 (#235)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| LMArena Expert | — | 1090 |
| ARC (AI2) Challenge | 92.2% | — |
| MMLU | 78.4% | — |
| TriviaQA | 80% | — |
Multilingual Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.2 3B: 26.2 (#281)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| LMArena Non-English | — | 1019 |
| LMArena Chinese | — | 1017 |
| LMArena German | — | 1056 |
| LMArena Russian | — | 949 |
Instruction Following Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.2 3B: 56.0 (#275)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| LMArena Instruction Following | — | 1089 |
Long Context Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.2 3B: 33.4 (#261)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| LMArena Longer Query | — | 1100 |
Writing & Preference Not comparable
DeepSeek-V2 (MoE-236B, May 2024): —, Llama 3.2 3B: 24.7 (#307)
| Benchmark | DeepSeek-V2 (MoE-236B, May 2024) | Llama 3.2 3B |
|---|---|---|
| LMArena Text | — | 1110 |
| LMArena Creative Writing | — | 1094 |
| EQ-Bench Creative Writing | — | 595 |
| LMArena Multi-Turn | — | 1105 |
Frequently asked questions
Is DeepSeek-V2 (MoE-236B, May 2024) better than Llama 3.2 3B?
Llama 3.2 3B has enough public results to be ranked (#321); DeepSeek-V2 (MoE-236B, May 2024) does not yet, so treat this comparison as directional.
Is DeepSeek-V2 (MoE-236B, May 2024) or Llama 3.2 3B better for coding?
DeepSeek-V2 (MoE-236B, May 2024) scores higher on coding benchmarks: 40.4 versus 27.6 in the Noometry coding category.
How many benchmarks do DeepSeek-V2 (MoE-236B, May 2024) and Llama 3.2 3B share?
2 benchmarks have published results for both models. DeepSeek-V2 (MoE-236B, May 2024) has 10 scored results on Noometry and Llama 3.2 3B has 18.