Model comparison
DeepSeek-R1 vs Llama 3.1-70B
DeepSeek-R1 is the stronger model overall, scoring 42.0 to 29.4 on the Noometry Index. Llama 3.1-70B costs 2.3× less per token, which makes it the better buy when DeepSeek-R1's lead doesn't matter for your workload.
Last verified . 29 shared benchmarks.
DeepSeek
MetaSummary
- They share 29 benchmarks with published results for both. DeepSeek-R1 scores higher in 8 categories and Llama 3.1-70B in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-R1 leads 42.0 to 12.7.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 66.4% for DeepSeek-R1 and 3.6% for Llama 3.1-70B.
- Llama 3.1-70B is cheaper at $0.40 / $0.40 per million input/output tokens, against $0.50 / $2.15 for DeepSeek-R1.
- DeepSeek-R1 accepts more context: 164K tokens versus 128K.
- Llama 3.1-70B has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-R1 | Llama 3.1-70B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 42.0 | 29.4 |
| Released | 2025-01-20 | 2024-07-23 |
| Weights | Proprietary | Open |
| Context window | 164K | 128K |
| Max output | 64K | 4K |
| Input $ / M tokens | $0.50 | $0.40 |
| Output $ / M tokens | $2.15 | $0.40 |
| Results tracked | 52 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-R1 leads
DeepSeek-R1: 45.9 (#63), Llama 3.1-70B: 29.8 (#292)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| WeirdML | 41.6% | 9% |
| LMArena Coding | 1427 | 1260 |
| Aider Polyglot | 71.4% | — |
| SciCode | 35.7% | — |
| BigCodeBench Instruct | — | 46.1% |
| LiveBench Coding | 66.7% | — |
| BigCodeBench Complete | — | 54.8% |
| ALE-Bench | 804.12 | — |
| AlgoTune | 1.7 | — |
Agentic & Tool Use DeepSeek-R1 leads
DeepSeek-R1: 30.4 (#68), Llama 3.1-70B: 24.9 (#104)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| BALROG | 34.9% | 27.9% |
| TheAgentCompany | — | 6.9% |
| DeepResearch Bench | 35.1% | — |
| METR Time Horizons | 53.8% | — |
Reasoning Llama 3.1-70B leads
DeepSeek-R1: 18.7 (#253), Llama 3.1-70B: 20.8 (#215)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| LMArena Hard Prompts | 1415 | 1241 |
| Epoch Capabilities Index | 141.3 | 125.91 |
| ARC-AGI-2 | 1.3% | — |
| SimpleBench | 40.8% | — |
| Kagi LLM Benchmark | 69.4% | — |
| ARC-AGI-1 | 21.2% | — |
| CritPt | 1.1% | — |
| LiveBench Reasoning | 83.2% | — |
| DTBench | — | 60% |
| LiveBench Data Analysis | 69.8% | — |
| LMCA | — | 14.8% |
| ForecastBench | 60 | — |
| LiveBench | 71.6% | — |
Math DeepSeek-R1 leads
DeepSeek-R1: 42.0 (#70), Llama 3.1-70B: 12.7 (#298)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 3.6% |
| Omni-MATH | 42.4% | 21% |
| LMArena Math | 1401 | 1252 |
| MATH Level 5 | 96.6% | 36.7% |
| LiveBench Math | 80.7% | — |
Knowledge DeepSeek-R1 leads
DeepSeek-R1: 44.3 (#81), Llama 3.1-70B: 24.1 (#264)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| GPQA Diamond | 76.3% | 44.2% |
| MMLU-Pro | 79.3% | 65.3% |
| GPQA (HELM) | 66.6% | 42.6% |
| LMArena Expert | 1394 | 1209 |
| Confabulations | 12.7% | — |
| Vectara Hallucination Rate | 11.3% | — |
| MMLU | — | 80.1% |
Multilingual DeepSeek-R1 leads
DeepSeek-R1: 52.6 (#80), Llama 3.1-70B: 38.9 (#220)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| LMArena Non-English | 1412 | 1219 |
| LMArena Chinese | 1443 | 1215 |
| LMArena French | 1415 | 1260 |
| LMArena German | 1405 | 1222 |
| LMArena Japanese | 1394 | 1132 |
| LMArena Korean | 1359 | 1140 |
| LMArena Russian | 1423 | 1234 |
| LMArena Spanish | 1410 | 1253 |
Instruction Following DeepSeek-R1 leads
DeepSeek-R1: 72.2 (#138), Llama 3.1-70B: 65.5 (#218)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| IFEval | 78.4% | 82.1% |
| LMArena Instruction Following | 1382 | 1231 |
| LiveBench Instruction Following | 80.5% | — |
Long Context DeepSeek-R1 leads
DeepSeek-R1: 45.4 (#36), Llama 3.1-70B: 37.6 (#208)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| LMArena Longer Query | 1391 | 1241 |
| Fiction.LiveBench | 75% | — |
Writing & Preference DeepSeek-R1 leads
DeepSeek-R1: 61.5 (#84), Llama 3.1-70B: 35.6 (#262)
| Benchmark | DeepSeek-R1 | Llama 3.1-70B |
|---|---|---|
| LMArena Text | 1428 | 1261 |
| LMArena Creative Writing | 1406 | 1232 |
| EQ-Bench Creative Writing | 1500 | 784 |
| WildBench | 82.8% | 75.8% |
| LMArena Multi-Turn | 1404 | 1256 |
| Short-Story Creative Writing | 83% | — |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than Llama 3.1-70B?
DeepSeek-R1 is the stronger model overall, scoring 42.0 to 29.4 on the Noometry Index. Llama 3.1-70B costs 2.3× less per token, which makes it the better buy when DeepSeek-R1's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-R1 or Llama 3.1-70B?
Llama 3.1-70B is cheaper. It lists at $0.40 per million input tokens and $0.40 per million output tokens; DeepSeek-R1 lists at $0.50 and $2.15.
Is DeepSeek-R1 or Llama 3.1-70B better for coding?
DeepSeek-R1 scores higher on coding benchmarks: 45.9 versus 29.8 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-R1 does, with 164K tokens against 128K.
How many benchmarks do DeepSeek-R1 and Llama 3.1-70B share?
29 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and Llama 3.1-70B has 35.