Model comparison
DeepSeek-V3 vs Llama-3.3-70B-Instruct
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 30.6 on the Noometry Index. Llama-3.3-70B-Instruct costs 2.6× less per token, which makes it the better buy when DeepSeek-V3's lead doesn't matter for your workload.
Last verified . 41 shared benchmarks.
DeepSeek
MetaSummary
- They share 41 benchmarks with published results for both. DeepSeek-V3 scores higher in 8 categories and Llama-3.3-70B-Instruct in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3 leads 32.1 to 15.3.
- The biggest single-benchmark swing is LiveBench Coding: 70.9% for DeepSeek-V3 and 36.6% for Llama-3.3-70B-Instruct.
- Llama-3.3-70B-Instruct is cheaper at $0.10 / $0.32 per million input/output tokens, against $0.24 / $0.90 for DeepSeek-V3.
- DeepSeek-V3 accepts more context: 164K tokens versus 128K.
Side by side
| DeepSeek-V3 | Llama-3.3-70B-Instruct | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 39.5 | 30.6 |
| Released | 2024-12-26 | 2024-12-06 |
| Weights | Open | Open |
| Context window | 164K | 128K |
| Max output | 164K | 4K |
| Input $ / M tokens | $0.24 | $0.10 |
| Output $ / M tokens | $0.90 | $0.32 |
| Results tracked | 60 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
DeepSeek-V3: 42.3 (#106), Llama-3.3-70B-Instruct: 31.0 (#290)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| SciCode | 35.8% | 26% |
| WeirdML | 36.1% | 14.4% |
| BigCodeBench Instruct | 50% | 46.9% |
| LiveBench Coding | 70.9% | 36.6% |
| LMArena Coding | 1368 | 1268 |
| BigCodeBench Complete | 62.2% | 57.5% |
| Aider Polyglot | 55.1% | — |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Llama-3.3-70B-Instruct: 25.8 (#105)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 31.9% |
| BALROG | — | 23% |
| METR Time Horizons | 49.6% | — |
Reasoning DeepSeek-V3 leads
DeepSeek-V3: 20.5 (#236), Llama-3.3-70B-Instruct: 14.1 (#327)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| SimpleBench | 27.2% | 19.9% |
| CritPt | 0% | 0% |
| LiveBench Reasoning | 65.8% | 50.8% |
| LMArena Hard Prompts | 1365 | 1257 |
| DTBench | 64.8% | 59.5% |
| LiveBench Data Analysis | 60.9% | 49.5% |
| LMCA | 15.5% | 17.5% |
| Epoch Capabilities Index | 135.94 | 127.33 |
| ForecastBench | 59.1 | 58.6 |
| LiveBench | 66.9% | 50.2% |
| Kagi LLM Benchmark | 52.3% | — |
| BIG-Bench Hard | 87.5% | — |
| HellaSwag | 88.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math DeepSeek-V3 leads
DeepSeek-V3: 32.1 (#219), Llama-3.3-70B-Instruct: 15.3 (#298)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 5.1% |
| LiveBench Math | 73.5% | 42.2% |
| LMArena Math | 1373 | 1267 |
| MATH Level 5 | 75.5% | 41.6% |
| Omni-MATH | 40.3% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
Knowledge DeepSeek-V3 leads
DeepSeek-V3: 37.5 (#155), Llama-3.3-70B-Instruct: 30.6 (#226)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| GPQA Diamond | 67.6% | 47.4% |
| Confabulations | 26.1% | 22.8% |
| Vectara Hallucination Rate | 6.1% | 4.1% |
| LMArena Expert | 1351 | 1225 |
| MMLU | 87.2% | 86.3% |
| MMLU-Pro | 72.3% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| TriviaQA | 82.9% | — |
Multilingual DeepSeek-V3 leads
DeepSeek-V3: 48.5 (#143), Llama-3.3-70B-Instruct: 39.9 (#220)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Non-English | 1358 | 1236 |
| LMArena Chinese | 1391 | 1217 |
| LMArena French | 1385 | 1281 |
| LMArena German | 1374 | 1251 |
| LMArena Japanese | 1333 | 1150 |
| LMArena Korean | 1319 | 1143 |
| LMArena Russian | 1373 | 1252 |
| LMArena Spanish | 1358 | 1270 |
Instruction Following DeepSeek-V3 leads
DeepSeek-V3: 72.8 (#130), Llama-3.3-70B-Instruct: 71.1 (#157)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LiveBench Instruction Following | 81.5% | 82.7% |
| LMArena Instruction Following | 1345 | 1242 |
| IFEval | 83.2% | — |
Long Context DeepSeek-V3 leads
DeepSeek-V3: 34.0 (#253), Llama-3.3-70B-Instruct: 26.4 (#295)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| Fiction.LiveBench | 50% | 33.3% |
| LMArena Longer Query | 1352 | 1256 |
Writing & Preference DeepSeek-V3 leads
DeepSeek-V3: 57.4 (#130), Llama-3.3-70B-Instruct: 47.6 (#207)
| Benchmark | DeepSeek-V3 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Text | 1375 | 1274 |
| LMArena Creative Writing | 1364 | 1250 |
| LMArena Multi-Turn | 1389 | 1280 |
| LiveBench Language | 49.1% | 39.2% |
| Short-Story Creative Writing | 77% | — |
| EQ-Bench Creative Writing | 1472 | — |
| WildBench | 83% | — |
Frequently asked questions
Is DeepSeek-V3 better than Llama-3.3-70B-Instruct?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 30.6 on the Noometry Index. Llama-3.3-70B-Instruct costs 2.6× less per token, which makes it the better buy when DeepSeek-V3's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3 or Llama-3.3-70B-Instruct?
Llama-3.3-70B-Instruct is cheaper. It lists at $0.10 per million input tokens and $0.32 per million output tokens; DeepSeek-V3 lists at $0.24 and $0.90.
Is DeepSeek-V3 or Llama-3.3-70B-Instruct better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 versus 31.0 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3 does, with 164K tokens against 128K.
How many benchmarks do DeepSeek-V3 and Llama-3.3-70B-Instruct share?
41 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Llama-3.3-70B-Instruct has 43.