Model comparison
Deepseek Coder v2 vs Llama 2-13B
Deepseek Coder v2 is the stronger model overall, scoring 35.3 to 29.1 on the Noometry Index.
Last verified . 20 shared benchmarks.
DeepSeek
MetaSummary
- They share 20 benchmarks with published results for both. Deepseek Coder v2 scores higher in 8 categories and Llama 2-13B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Deepseek Coder v2 leads 22.6 to 12.3.
Side by side
| Deepseek Coder v2 | Llama 2-13B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 35.3 | 29.1 |
| Released | 2024-06-17 | 2023-07-18 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 24 | 32 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Deepseek Coder v2 leads
Deepseek Coder v2: 37.5 (#178), Llama 2-13B: 30.4 (#287)
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| LMArena Coding | 1251 | 1062 |
| BigCodeBench Instruct | 48.2% | — |
| BigCodeBench Complete | 59.7% | — |
| HumanEval+ | 82.3% | — |
| MBPP+ | 75.1% | — |
Reasoning Deepseek Coder v2 leads
Deepseek Coder v2: 22.6 (#175), Llama 2-13B: 12.3 (#333)
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| LMArena Hard Prompts | 1207 | 1051 |
| WinoGrande | 83.7% | 72.8% |
| Chess Puzzles | — | 0% |
| DTBench | — | 42.2% |
| BIG-Bench Hard | — | 58.2% |
| Epoch Capabilities Index | — | 105.8 |
| HellaSwag | — | 80.7% |
| LAMBADA | — | 76.5% |
| PIQA | — | 80.8% |
Math Deepseek Coder v2 leads
Deepseek Coder v2: 32.3 (#188), Llama 2-13B: 28.7 (#226)
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| LMArena Math | 1241 | 1065 |
| GSM8K | 94.5% | 36.9% |
Knowledge Deepseek Coder v2 leads
Deepseek Coder v2: 31.7 (#209), Llama 2-13B: 27.5 (#244)
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| LMArena Expert | 1181 | 1030 |
| ARC (AI2) Challenge | 64.3% | 60.3% |
| BoolQ | — | 82.4% |
| MMLU | — | 55.6% |
| OpenBookQA | — | 57% |
| TriviaQA | — | 79.6% |
Multimodal Not comparable
Deepseek Coder v2: —, Llama 2-13B: —
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| ScienceQA | — | 55.8% |
Multilingual Deepseek Coder v2 leads
Deepseek Coder v2: 36.4 (#235), Llama 2-13B: 26.7 (#274)
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| LMArena Non-English | 1182 | 1024 |
| LMArena Chinese | 1201 | 1001 |
| LMArena French | 1184 | 1043 |
| LMArena German | 1164 | 1009 |
| LMArena Japanese | 1126 | 894 |
| LMArena Korean | 1104 | 953 |
| LMArena Russian | 1188 | 1055 |
| LMArena Spanish | 1153 | 1086 |
Instruction Following Deepseek Coder v2 leads
Deepseek Coder v2: 61.9 (#237), Llama 2-13B: 53.5 (#282)
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| LMArena Instruction Following | 1180 | 1046 |
Long Context Deepseek Coder v2 leads
Deepseek Coder v2: 37.0 (#219), Llama 2-13B: 32.3 (#264)
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| LMArena Longer Query | 1219 | 1064 |
Writing & Preference Deepseek Coder v2 leads
Deepseek Coder v2: 38.4 (#249), Llama 2-13B: 30.0 (#284)
| Benchmark | Deepseek Coder v2 | Llama 2-13B |
|---|---|---|
| LMArena Text | 1191 | 1085 |
| LMArena Creative Writing | 1120 | 1047 |
| LMArena Multi-Turn | 1176 | 1050 |
Frequently asked questions
Is Deepseek Coder v2 better than Llama 2-13B?
Deepseek Coder v2 is the stronger model overall, scoring 35.3 to 29.1 on the Noometry Index.
Is Deepseek Coder v2 or Llama 2-13B better for coding?
Deepseek Coder v2 scores higher on coding benchmarks: 37.5 versus 30.4 in the Noometry coding category.
How many benchmarks do Deepseek Coder v2 and Llama 2-13B share?
20 benchmarks have published results for both models. Deepseek Coder v2 has 24 scored results on Noometry and Llama 2-13B has 32.