Model comparison
DeepSeek-V3.1 vs Llama-3.3-70B-Instruct
DeepSeek-V3.1 is the stronger model overall, scoring 42.8 to 30.6 on the Noometry Index. Llama-3.3-70B-Instruct costs 2.7× less per token, which makes it the better buy when DeepSeek-V3.1's lead doesn't matter for your workload.
Last verified . 25 shared benchmarks.
DeepSeek
MetaSummary
- They share 25 benchmarks with published results for both. DeepSeek-V3.1 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-V3.1 leads 38.9 to 15.3.
- The biggest single-benchmark swing is WeirdML: 38.4% for DeepSeek-V3.1 and 14.4% for Llama-3.3-70B-Instruct.
- Llama-3.3-70B-Instruct is cheaper at $0.10 / $0.32 per million input/output tokens, against $0.25 / $0.95 for DeepSeek-V3.1.
- DeepSeek-V3.1 accepts more context: 164K tokens versus 128K.
Side by side
| DeepSeek-V3.1 | Llama-3.3-70B-Instruct | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 42.8 | 30.6 |
| Released | 2025-08-21 | 2024-12-06 |
| Weights | Open | Open |
| Context window | 164K | 128K |
| Max output | 8K | 4K |
| Input $ / M tokens | $0.25 | $0.10 |
| Output $ / M tokens | $0.95 | $0.32 |
| Results tracked | 27 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.1 leads
DeepSeek-V3.1: 40.3 (#144), Llama-3.3-70B-Instruct: 31.0 (#290)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| WeirdML | 38.4% | 14.4% |
| LMArena Coding | 1417 | 1268 |
| SciCode | — | 26% |
| BigCodeBench Instruct | — | 46.9% |
| LiveBench Coding | — | 36.6% |
| BigCodeBench Complete | — | 57.5% |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, Llama-3.3-70B-Instruct: 25.8 (#105)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 31.9% |
| BALROG | — | 23% |
Reasoning DeepSeek-V3.1 leads
DeepSeek-V3.1: 27.9 (#110), Llama-3.3-70B-Instruct: 14.1 (#327)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| SimpleBench | 40% | 19.9% |
| LMArena Hard Prompts | 1417 | 1257 |
| DTBench | 82.7% | 59.5% |
| LMCA | 24.3% | 17.5% |
| Epoch Capabilities Index | 139.92 | 127.33 |
| ForecastBench | 58 | 58.6 |
| Kagi LLM Benchmark | 53.2% | — |
| CritPt | — | 0% |
| LiveBench Reasoning | — | 50.8% |
| LiveBench Data Analysis | — | 49.5% |
| LiveBench | — | 50.2% |
Math DeepSeek-V3.1 leads
DeepSeek-V3.1: 38.9 (#122), Llama-3.3-70B-Instruct: 15.3 (#298)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Math | 1420 | 1267 |
| OTIS Mock AIME 2024-2025 | — | 5.1% |
| LiveBench Math | — | 42.2% |
| MATH Level 5 | — | 41.6% |
Knowledge DeepSeek-V3.1 leads
DeepSeek-V3.1: 43.7 (#90), Llama-3.3-70B-Instruct: 30.6 (#226)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| Vectara Hallucination Rate | 5.5% | 4.1% |
| LMArena Expert | 1405 | 1225 |
| GPQA Diamond | — | 47.4% |
| Confabulations | — | 22.8% |
| MMLU | — | 86.3% |
Multilingual DeepSeek-V3.1 leads
DeepSeek-V3.1: 51.6 (#106), Llama-3.3-70B-Instruct: 39.9 (#220)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Non-English | 1400 | 1236 |
| LMArena Chinese | 1469 | 1217 |
| LMArena French | 1447 | 1281 |
| LMArena German | 1411 | 1251 |
| LMArena Japanese | 1378 | 1150 |
| LMArena Korean | 1337 | 1143 |
| LMArena Russian | 1405 | 1252 |
| LMArena Spanish | 1431 | 1270 |
Instruction Following DeepSeek-V3.1 leads
DeepSeek-V3.1: 73.9 (#110), Llama-3.3-70B-Instruct: 71.1 (#157)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Instruction Following | 1400 | 1242 |
| LiveBench Instruction Following | — | 82.7% |
Long Context DeepSeek-V3.1 leads
DeepSeek-V3.1: 36.3 (#232), Llama-3.3-70B-Instruct: 26.4 (#295)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| Fiction.LiveBench | 52.8% | 33.3% |
| LMArena Longer Query | 1422 | 1256 |
Writing & Preference DeepSeek-V3.1 leads
DeepSeek-V3.1: 60.3 (#98), Llama-3.3-70B-Instruct: 47.6 (#207)
| Benchmark | DeepSeek-V3.1 | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Text | 1420 | 1274 |
| LMArena Creative Writing | 1401 | 1250 |
| LMArena Multi-Turn | 1408 | 1280 |
| EQ-Bench Creative Writing | 1436 | — |
| LiveBench Language | — | 39.2% |
Frequently asked questions
Is DeepSeek-V3.1 better than Llama-3.3-70B-Instruct?
DeepSeek-V3.1 is the stronger model overall, scoring 42.8 to 30.6 on the Noometry Index. Llama-3.3-70B-Instruct costs 2.7× less per token, which makes it the better buy when DeepSeek-V3.1's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.1 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-V3.1 lists at $0.25 and $0.95.
Is DeepSeek-V3.1 or Llama-3.3-70B-Instruct better for coding?
DeepSeek-V3.1 scores higher on coding benchmarks: 40.3 versus 31.0 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3.1 does, with 164K tokens against 128K.
How many benchmarks do DeepSeek-V3.1 and Llama-3.3-70B-Instruct share?
25 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and Llama-3.3-70B-Instruct has 43.