Model comparison
DeepSeek-V3 vs Llama 3.2 90B
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 27.5 on the Noometry Index.
Last verified . 5 shared benchmarks.
DeepSeek
MetaSummary
- They share 5 benchmarks with published results for both. DeepSeek-V3 scores higher in 2 categories and Llama 3.2 90B in 1 category; 3 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3 leads 32.1 to 11.1.
- The biggest single-benchmark swing is MATH Level 5: 75.5% for DeepSeek-V3 and 39.4% for Llama 3.2 90B.
Side by side
| DeepSeek-V3 | Llama 3.2 90B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 39.5 | 27.5 |
| Released | 2024-12-26 | 2024-09-24 |
| Weights | Open | Open |
| Context window | 164K | — |
| Max output | 164K | — |
| Input $ / M tokens | $0.24 | — |
| Output $ / M tokens | $0.90 | — |
| Results tracked | 60 | 9 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Not comparable
DeepSeek-V3: 42.3 (#106), Llama 3.2 90B: —
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| Aider Polyglot | 55.1% | — |
| SciCode | 35.8% | — |
| WeirdML | 36.1% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| LMArena Coding | 1368 | — |
| BigCodeBench Complete | 62.2% | — |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Llama 3.2 90B: 30.0 (#80)
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| BALROG | — | 27.3% |
| METR Time Horizons | 49.6% | — |
Reasoning Llama 3.2 90B leads
DeepSeek-V3: 20.5 (#236), Llama 3.2 90B: 21.7 (#217)
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| Epoch Capabilities Index | 135.94 | 125.5 |
| SimpleBench | 27.2% | — |
| Kagi LLM Benchmark | 52.3% | — |
| CritPt | 0% | — |
| EnigmaEval | — | 0.4% |
| LiveBench Reasoning | 65.8% | — |
| LMArena Hard Prompts | 1365 | — |
| DTBench | 64.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| LMCA | 15.5% | — |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math DeepSeek-V3 leads
DeepSeek-V3: 32.1 (#219), Llama 3.2 90B: 11.1 (#308)
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 2.6% |
| MATH Level 5 | 75.5% | 39.4% |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| LMArena Math | 1373 | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
Knowledge DeepSeek-V3 leads
DeepSeek-V3: 37.5 (#155), Llama 3.2 90B: 21.7 (#274)
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| GPQA Diamond | 67.6% | 41% |
| MMLU | 87.2% | 80.3% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| LMArena Expert | 1351 | — |
| ARC (AI2) Challenge | 95.3% | — |
| TriviaQA | 82.9% | — |
Multimodal Not comparable
DeepSeek-V3: —, Llama 3.2 90B: 25.4 (#124)
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| LMArena Vision | — | 1000 |
| GeoBench | — | 52% |
Multilingual Not comparable
DeepSeek-V3: 48.5 (#143), Llama 3.2 90B: —
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| LMArena Non-English | 1358 | — |
| LMArena Chinese | 1391 | — |
| LMArena French | 1385 | — |
| LMArena German | 1374 | — |
| LMArena Japanese | 1333 | — |
| LMArena Korean | 1319 | — |
| LMArena Russian | 1373 | — |
| LMArena Spanish | 1358 | — |
Instruction Following Not comparable
DeepSeek-V3: 72.8 (#130), Llama 3.2 90B: —
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |
| LMArena Instruction Following | 1345 | — |
Long Context Not comparable
DeepSeek-V3: 34.0 (#253), Llama 3.2 90B: —
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| Fiction.LiveBench | 50% | — |
| LMArena Longer Query | 1352 | — |
Writing & Preference Not comparable
DeepSeek-V3: 57.4 (#130), Llama 3.2 90B: —
| Benchmark | DeepSeek-V3 | Llama 3.2 90B |
|---|---|---|
| LMArena Text | 1375 | — |
| LMArena Creative Writing | 1364 | — |
| Short-Story Creative Writing | 77% | — |
| EQ-Bench Creative Writing | 1472 | — |
| WildBench | 83% | — |
| LMArena Multi-Turn | 1389 | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than Llama 3.2 90B?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 27.5 on the Noometry Index.
How many benchmarks do DeepSeek-V3 and Llama 3.2 90B share?
5 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Llama 3.2 90B has 9.