Model comparison
Llama 3.2 3B vs Llama-3.3-70B-Instruct
Llama-3.3-70B-Instruct is the stronger model overall, scoring 30.6 to 28.9 on the Noometry Index.
Last verified . 17 shared benchmarks.
MetaSummary
- They share 17 benchmarks with published results for both. Llama 3.2 3B scores higher in 3 categories and Llama-3.3-70B-Instruct in 6 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Llama-3.3-70B-Instruct leads 47.6 to 24.7.
- The biggest single-benchmark swing is BigCodeBench Complete: 28.3% for Llama 3.2 3B and 57.5% for Llama-3.3-70B-Instruct.
- Llama 3.2 3B is cheaper at $0.05 / $0.33 per million input/output tokens, against $0.10 / $0.32 for Llama-3.3-70B-Instruct.
- Llama 3.2 3B accepts more context: 131K tokens versus 128K.
Side by side
| Llama 3.2 3B | Llama-3.3-70B-Instruct | |
|---|---|---|
| Provider | Meta | Meta |
| Noometry Index | 28.9 | 30.6 |
| Released | 2024-09-24 | 2024-12-06 |
| Weights | Open | Open |
| Context window | 131K | 128K |
| Max output | 118K | 4K |
| Input $ / M tokens | $0.05 | $0.10 |
| Output $ / M tokens | $0.33 | $0.32 |
| Results tracked | 18 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Llama-3.3-70B-Instruct leads
Llama 3.2 3B: 27.6 (#319), Llama-3.3-70B-Instruct: 31.0 (#290)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| BigCodeBench Instruct | 23.4% | 46.9% |
| LMArena Coding | 1098 | 1268 |
| BigCodeBench Complete | 28.3% | 57.5% |
| SciCode | — | 26% |
| WeirdML | — | 14.4% |
| LiveBench Coding | — | 36.6% |
Agentic & Tool Use Llama-3.3-70B-Instruct leads
Llama 3.2 3B: 20.1 (#143), Llama-3.3-70B-Instruct: 25.8 (#105)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| Berkeley Function Calling Leaderboard | 21.9% | 31.9% |
| BALROG | 10.1% | 23% |
Reasoning Llama 3.2 3B leads
Llama 3.2 3B: 21.0 (#228), Llama-3.3-70B-Instruct: 14.1 (#327)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Hard Prompts | 1095 | 1257 |
| SimpleBench | — | 19.9% |
| CritPt | — | 0% |
| LiveBench Reasoning | — | 50.8% |
| DTBench | — | 59.5% |
| LiveBench Data Analysis | — | 49.5% |
| LMCA | — | 17.5% |
| Epoch Capabilities Index | — | 127.33 |
| ForecastBench | — | 58.6 |
| LiveBench | — | 50.2% |
Math Llama 3.2 3B leads
Llama 3.2 3B: 32.4 (#214), Llama-3.3-70B-Instruct: 15.3 (#298)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Math | 1126 | 1267 |
| OTIS Mock AIME 2024-2025 | — | 5.1% |
| LiveBench Math | — | 42.2% |
| MATH Level 5 | — | 41.6% |
Knowledge Too close to call
Llama 3.2 3B: 29.7 (#235), Llama-3.3-70B-Instruct: 30.6 (#226)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Expert | 1090 | 1225 |
| GPQA Diamond | — | 47.4% |
| Confabulations | — | 22.8% |
| Vectara Hallucination Rate | — | 4.1% |
| MMLU | — | 86.3% |
Multilingual Llama-3.3-70B-Instruct leads
Llama 3.2 3B: 26.2 (#281), Llama-3.3-70B-Instruct: 39.9 (#220)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Non-English | 1019 | 1236 |
| LMArena Chinese | 1017 | 1217 |
| LMArena German | 1056 | 1251 |
| LMArena Russian | 949 | 1252 |
| LMArena French | — | 1281 |
| LMArena Japanese | — | 1150 |
| LMArena Korean | — | 1143 |
| LMArena Spanish | — | 1270 |
Instruction Following Llama-3.3-70B-Instruct leads
Llama 3.2 3B: 56.0 (#275), Llama-3.3-70B-Instruct: 71.1 (#157)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Instruction Following | 1089 | 1242 |
| LiveBench Instruction Following | — | 82.7% |
Long Context Llama 3.2 3B leads
Llama 3.2 3B: 33.4 (#261), Llama-3.3-70B-Instruct: 26.4 (#295)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Longer Query | 1100 | 1256 |
| Fiction.LiveBench | — | 33.3% |
Writing & Preference Llama-3.3-70B-Instruct leads
Llama 3.2 3B: 24.7 (#307), Llama-3.3-70B-Instruct: 47.6 (#207)
| Benchmark | Llama 3.2 3B | Llama-3.3-70B-Instruct |
|---|---|---|
| LMArena Text | 1110 | 1274 |
| LMArena Creative Writing | 1094 | 1250 |
| LMArena Multi-Turn | 1105 | 1280 |
| EQ-Bench Creative Writing | 595 | — |
| LiveBench Language | — | 39.2% |
Frequently asked questions
Is Llama 3.2 3B better than Llama-3.3-70B-Instruct?
Llama-3.3-70B-Instruct is the stronger model overall, scoring 30.6 to 28.9 on the Noometry Index.
Which is cheaper, Llama 3.2 3B or Llama-3.3-70B-Instruct?
Llama 3.2 3B is cheaper. It lists at $0.05 per million input tokens and $0.33 per million output tokens; Llama-3.3-70B-Instruct lists at $0.10 and $0.32.
Is Llama 3.2 3B or Llama-3.3-70B-Instruct better for coding?
Llama-3.3-70B-Instruct scores higher on coding benchmarks: 31.0 versus 27.6 in the Noometry coding category.
Which has the bigger context window?
Llama 3.2 3B does, with 131K tokens against 128K.
How many benchmarks do Llama 3.2 3B and Llama-3.3-70B-Instruct share?
17 benchmarks have published results for both models. Llama 3.2 3B has 18 scored results on Noometry and Llama-3.3-70B-Instruct has 43.