Model comparison
DeepSeek-R1 vs Llama 3.2 3B
DeepSeek-R1 is the stronger model overall, scoring 42.3 to 28.9 on the Noometry Index. Llama 3.2 3B costs 7.6× less per token, which makes it the better buy when DeepSeek-R1's lead doesn't matter for your workload.
Last verified . 15 shared benchmarks.
DeepSeek
MetaSummary
- They share 15 benchmarks with published results for both. DeepSeek-R1 scores higher in 8 categories and Llama 3.2 3B in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-R1 leads 61.4 to 24.7.
- The biggest single-benchmark swing is BALROG: 34.9% for DeepSeek-R1 and 10.1% for Llama 3.2 3B.
- Llama 3.2 3B is cheaper at $0.05 / $0.33 per million input/output tokens, against $0.50 / $2.15 for DeepSeek-R1.
- DeepSeek-R1 accepts more context: 164K tokens versus 131K.
- Llama 3.2 3B has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-R1 | Llama 3.2 3B | |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 42.3 | 28.9 |
| Released | 2025-01-20 | 2024-09-24 |
| Weights | Proprietary | Open |
| Context window | 164K | 131K |
| Max output | 64K | 118K |
| Input $ / M tokens | $0.50 | $0.05 |
| Output $ / M tokens | $2.15 | $0.33 |
| Results tracked | 52 | 18 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-R1 leads
DeepSeek-R1: 46.3 (#68), Llama 3.2 3B: 27.6 (#319)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| LMArena Coding | 1427 | 1098 |
| Aider Polyglot | 71.4% | — |
| SciCode | 35.7% | — |
| WeirdML | 41.6% | — |
| BigCodeBench Instruct | — | 23.4% |
| LiveBench Coding | 66.7% | — |
| BigCodeBench Complete | — | 28.3% |
| ALE-Bench | 804.12 | — |
| AlgoTune | 1.7 | — |
Agentic & Tool Use DeepSeek-R1 leads
DeepSeek-R1: 30.7 (#75), Llama 3.2 3B: 20.1 (#143)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| BALROG | 34.9% | 10.1% |
| Berkeley Function Calling Leaderboard | — | 21.9% |
| DeepResearch Bench | 35.1% | — |
| METR Time Horizons | 53.8% | — |
Reasoning Llama 3.2 3B leads
DeepSeek-R1: 18.6 (#278), Llama 3.2 3B: 21.0 (#228)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| LMArena Hard Prompts | 1416 | 1095 |
| 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% | — |
| LiveBench Data Analysis | 69.8% | — |
| Epoch Capabilities Index | 141.29 | — |
| ForecastBench | 60 | — |
| LiveBench | 71.6% | — |
Math DeepSeek-R1 leads
DeepSeek-R1: 43.8 (#79), Llama 3.2 3B: 32.4 (#214)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| LMArena Math | 1400 | 1126 |
| OTIS Mock AIME 2024-2025 | 66.4% | — |
| Omni-MATH | 42.4% | — |
| LiveBench Math | 80.7% | — |
| MATH Level 5 | 96.6% | — |
Knowledge DeepSeek-R1 leads
DeepSeek-R1: 44.5 (#87), Llama 3.2 3B: 29.7 (#235)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| LMArena Expert | 1394 | 1090 |
| GPQA Diamond | 76.3% | — |
| MMLU-Pro | 79.3% | — |
| Confabulations | 12.7% | — |
| Vectara Hallucination Rate | 11.3% | — |
| GPQA (HELM) | 66.6% | — |
Multilingual DeepSeek-R1 leads
DeepSeek-R1: 52.4 (#85), Llama 3.2 3B: 26.2 (#281)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| LMArena Non-English | 1412 | 1019 |
| LMArena Chinese | 1442 | 1017 |
| LMArena German | 1404 | 1056 |
| LMArena Russian | 1423 | 949 |
| LMArena French | 1417 | — |
| LMArena Japanese | 1391 | — |
| LMArena Korean | 1360 | — |
| LMArena Spanish | 1411 | — |
Instruction Following DeepSeek-R1 leads
DeepSeek-R1: 72.0 (#143), Llama 3.2 3B: 56.0 (#275)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| LMArena Instruction Following | 1382 | 1089 |
| LiveBench Instruction Following | 80.5% | — |
| IFEval | 78.4% | — |
Long Context DeepSeek-R1 leads
DeepSeek-R1: 45.4 (#36), Llama 3.2 3B: 33.4 (#261)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| LMArena Longer Query | 1391 | 1100 |
| Fiction.LiveBench | 75% | — |
Writing & Preference DeepSeek-R1 leads
DeepSeek-R1: 61.4 (#88), Llama 3.2 3B: 24.7 (#307)
| Benchmark | DeepSeek-R1 | Llama 3.2 3B |
|---|---|---|
| LMArena Text | 1428 | 1110 |
| LMArena Creative Writing | 1405 | 1094 |
| EQ-Bench Creative Writing | 1500 | 595 |
| LMArena Multi-Turn | 1405 | 1105 |
| Short-Story Creative Writing | 83% | — |
| WildBench | 82.8% | — |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than Llama 3.2 3B?
DeepSeek-R1 is the stronger model overall, scoring 42.3 to 28.9 on the Noometry Index. Llama 3.2 3B costs 7.6× 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 3.2 3B?
Llama 3.2 3B is cheaper. It lists at $0.05 per million input tokens and $0.33 per million output tokens; DeepSeek-R1 lists at $0.50 and $2.15.
Is DeepSeek-R1 or Llama 3.2 3B better for coding?
DeepSeek-R1 scores higher on coding benchmarks: 46.3 versus 27.6 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-R1 does, with 164K tokens against 131K.
How many benchmarks do DeepSeek-R1 and Llama 3.2 3B share?
15 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and Llama 3.2 3B has 18.