Model comparison
DeepSeek-V2.5 (Sep 2024) vs Llama-3.3-70B-Instruct
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.0 to 30.3 on the Noometry Index.
Last verified . 19 shared benchmarks.
DeepSeek
MetaSummary
- They share 19 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 7 categories and Llama-3.3-70B-Instruct in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V2.5 (Sep 2024) leads 33.3 to 14.3.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 37.0 | 30.3 |
| Released | 2024-09-06 | 2024-12-06 |
| Weights | Open | Open |
| Context window | — | 128K |
| Max output | — | 4K |
| Input $ / M tokens | — | $0.10 |
| Output $ / M tokens | — | $0.32 |
| Results tracked | 22 | 43 |
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): 31.3 (#274), Llama-3.3-70B-Instruct: 30.6 (#284)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| BigCodeBench Instruct | 48.6% | 46.9% |
| LMArena Coding | 1310 | 1268 |
| BigCodeBench Complete | 53.2% | 57.5% |
| Aider Polyglot | 17.8% | — |
| SciCode | — | 26% |
| WeirdML | — | 14.4% |
| LiveBench Coding | — | 36.6% |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Llama-3.3-70B-Instruct: 25.6 (#100)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 31.9% |
| BALROG | — | 23% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 24.5 (#145), Llama-3.3-70B-Instruct: 13.7 (#325)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1257 |
| SimpleBench | — | 19.9% |
| CritPt | — | 0% |
| LiveBench Reasoning | — | 50.8% |
| DTBench | — | 59.5% |
| LiveBench Data Analysis | — | 49.5% |
| LMCA | — | 17.5% |
| Epoch Capabilities Index | — | 127.32 |
| ForecastBench | — | 58.6 |
| LiveBench | — | 50.2% |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 33.3 (#177), Llama-3.3-70B-Instruct: 14.3 (#292)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Math | 1289 | 1268 |
| OTIS Mock AIME 2024-2025 | — | 5.1% |
| LiveBench Math | — | 42.2% |
| MATH Level 5 | — | 41.6% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.2 (#190), Llama-3.3-70B-Instruct: 30.5 (#220)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Expert | 1266 | 1225 |
| GPQA Diamond | — | 47.4% |
| Confabulations | — | 22.8% |
| Vectara Hallucination Rate | — | 4.1% |
| MMLU | — | 86.3% |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 42.7 (#188), Llama-3.3-70B-Instruct: 40.1 (#215)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Non-English | 1273 | 1236 |
| LMArena Chinese | 1317 | 1217 |
| LMArena French | 1288 | 1280 |
| LMArena German | 1258 | 1251 |
| LMArena Japanese | 1228 | 1149 |
| LMArena Korean | 1209 | 1142 |
| LMArena Russian | 1289 | 1252 |
| LMArena Spanish | 1248 | 1269 |
Instruction Following Llama-3.3-70B-Instruct leads
DeepSeek-V2.5 (Sep 2024): 67.7 (#189), Llama-3.3-70B-Instruct: 71.3 (#152)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Instruction Following | 1280 | 1242 |
| LiveBench Instruction Following | — | 82.7% |
Long Context DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#169), Llama-3.3-70B-Instruct: 26.4 (#290)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Longer Query | 1301 | 1256 |
| Fiction.LiveBench | — | 33.3% |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 50.0 (#182), Llama-3.3-70B-Instruct: 47.8 (#203)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Text | 1294 | 1274 |
| LMArena Creative Writing | 1285 | 1250 |
| LMArena Multi-Turn | 1297 | 1279 |
| LiveBench Language | — | 39.2% |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Llama-3.3-70B-Instruct?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.0 to 30.3 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Llama-3.3-70B-Instruct better for coding?
They score almost the same on coding (31.3 vs 30.6); test both on your own repository before choosing.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Llama-3.3-70B-Instruct share?
19 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Llama-3.3-70B-Instruct has 43.