Model comparison
DeepSeek-V2.5 (Sep 2024) vs Llama 3-8B
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 25.5 on the Noometry Index.
Last verified . 21 shared benchmarks.
DeepSeek
MetaSummary
- They share 21 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 8 categories and Llama 3-8B in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V2.5 (Sep 2024) leads 35.9 to 8.8.
- The biggest single-benchmark swing is BigCodeBench Instruct: 48.6% for DeepSeek-V2.5 (Sep 2024) and 31.9% for Llama 3-8B.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Llama 3-8B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 37.6 | 25.5 |
| Released | 2024-09-06 | 2024-04-18 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 22 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Llama 3-8B: 31.0 (#289)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama 3-8B |
|---|---|---|
| BigCodeBench Instruct | 48.6% | 31.9% |
| LMArena Coding | 1309 | 1152 |
| BigCodeBench Complete | 53.2% | 36.9% |
| HumanEval+ | 83.5% | 56.7% |
| MBPP+ | 74.1% | 54.8% |
| Aider Polyglot | 17.8% | — |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Llama 3-8B: 14.3 (#326)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama 3-8B |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1133 |
| Chess Puzzles | — | 0% |
| DTBench | — | 43.9% |
| Adversarial NLI | — | 57.3% |
| Epoch Capabilities Index | — | 116.45 |
| ForecastBench | — | 58.6 |
| WinoGrande | — | 75.7% |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Llama 3-8B: 8.8 (#323)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama 3-8B |
|---|---|---|
| LMArena Math | 1288 | 1151 |
| OTIS Mock AIME 2024-2025 | — | 1.9% |
| MATH Level 5 | — | 6.1% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Llama 3-8B: 7.8 (#308)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama 3-8B |
|---|---|---|
| LMArena Expert | 1266 | 1113 |
| GPQA Diamond | — | 26.1% |
| ARC (AI2) Challenge | — | 82.8% |
| MMLU | — | 68.8% |
| OpenBookQA | — | 82.6% |
| TriviaQA | — | 67.7% |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Llama 3-8B: 30.8 (#261)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama 3-8B |
|---|---|---|
| LMArena Non-English | 1273 | 1098 |
| LMArena Chinese | 1318 | 1076 |
| LMArena French | 1289 | 1159 |
| LMArena German | 1258 | 1104 |
| LMArena Japanese | 1228 | 967 |
| LMArena Korean | 1209 | 1004 |
| LMArena Russian | 1289 | 1109 |
| LMArena Spanish | 1248 | 1173 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Llama 3-8B: 58.4 (#260)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama 3-8B |
|---|---|---|
| LMArena Instruction Following | 1280 | 1127 |
Long Context DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Llama 3-8B: 34.2 (#251)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama 3-8B |
|---|---|---|
| LMArena Longer Query | 1301 | 1128 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Llama 3-8B: 37.5 (#256)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama 3-8B |
|---|---|---|
| LMArena Text | 1294 | 1166 |
| LMArena Creative Writing | 1285 | 1150 |
| LMArena Multi-Turn | 1297 | 1152 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Llama 3-8B?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 25.5 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Llama 3-8B better for coding?
They score almost the same on coding (31.7 vs 31.0); test both on your own repository before choosing.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Llama 3-8B share?
21 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Llama 3-8B has 34.