Model comparison
GPT-5 Nano vs Llama 3.2 3B
GPT-5 Nano is the stronger model overall, scoring 33.4 to 28.4 on the Noometry Index.
Last verified . 15 shared benchmarks.
MetaSummary
- They share 15 benchmarks with published results for both. GPT-5 Nano scores higher in 6 categories and Llama 3.2 3B in 3 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in multilingual, where GPT-5 Nano leads 45.5 to 26.4.
- The biggest single-benchmark swing is Berkeley Function Calling Leaderboard: 51.5% for GPT-5 Nano and 21.9% for Llama 3.2 3B.
- Llama 3.2 3B is cheaper at $0.05 / $0.33 per million input/output tokens, against $0.05 / $0.40 for GPT-5 Nano.
- GPT-5 Nano accepts more context: 400K tokens versus 131K.
- Llama 3.2 3B has downloadable open weights; the other is API-only.
Side by side
| GPT-5 Nano | Llama 3.2 3B | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 33.4 | 28.4 |
| Released | 2025-08-07 | 2024-09-24 |
| Weights | Proprietary | Open |
| Context window | 400K | 131K |
| Max output | 128K | 118K |
| Input $ / M tokens | $0.05 | $0.05 |
| Output $ / M tokens | $0.40 | $0.33 |
| Results tracked | 48 | 18 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 Nano leads
GPT-5 Nano: 33.3 (#248), Llama 3.2 3B: 27.1 (#314)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Coding | 1351 | 1098 |
| SWE-bench Verified (bash only) | 34.8% | — |
| WeirdML | 38.1% | — |
| BigCodeBench Instruct | — | 23.4% |
| BigCodeBench Complete | — | 28.3% |
| ALE-Bench | 718.67 | — |
Agentic & Tool Use GPT-5 Nano leads
GPT-5 Nano: 25.6 (#99), Llama 3.2 3B: 19.9 (#135)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 51.5% | 21.9% |
| Terminal-Bench | 21.8% | — |
| BALROG | — | 10.1% |
Reasoning Llama 3.2 3B leads
GPT-5 Nano: 16.4 (#291), Llama 3.2 3B: 20.1 (#225)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Hard Prompts | 1329 | 1095 |
| ARC-AGI-2 | 2.6% | — |
| Kagi LLM Benchmark | 62.2% | — |
| ARC-AGI-1 | 20.7% | — |
| Chess Puzzles | 27% | — |
| Mystery Game Puzzles | 9% | — |
| DTBench | 62.7% | — |
| LMCA | 7.9% | — |
| Epoch Capabilities Index | 139.39 | — |
| ForecastBench | 59.1 | — |
Math Too close to call
GPT-5 Nano: 29.1 (#221), Llama 3.2 3B: 29.9 (#213)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Math | 1318 | 1126 |
| FrontierMath (Tiers 1-3) | 20% | — |
| FrontierMath Tier 4 | 2.4% | — |
| OTIS Mock AIME 2024-2025 | 81.1% | — |
| ProofBench | 12% | — |
| Omni-MATH | 54.6% | — |
| MATH Level 5 | 95.2% | — |
| FrontierMath (Feb 2025 set) | 8.3% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge GPT-5 Nano leads
GPT-5 Nano: 35.8 (#166), Llama 3.2 3B: 29.1 (#232)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Expert | 1321 | 1090 |
| GPQA Diamond | 69.4% | — |
| SimpleQA Verified | 11.7% | — |
| MMLU-Pro | 77.8% | — |
| Vectara Hallucination Rate | 10.5% | — |
| GPQA (HELM) | 67.9% | — |
Multimodal Not comparable
GPT-5 Nano: 30.7 (#102), Llama 3.2 3B: —
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Vision | 1159 | — |
| VPCT | 37.2% | — |
Multilingual GPT-5 Nano leads
GPT-5 Nano: 45.5 (#167), Llama 3.2 3B: 26.4 (#276)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Non-English | 1313 | 1019 |
| LMArena Chinese | 1355 | 1017 |
| LMArena German | 1329 | 1056 |
| LMArena Russian | 1296 | 949 |
| LMArena Korean | 1269 | — |
| LMArena Spanish | 1360 | — |
Instruction Following GPT-5 Nano leads
GPT-5 Nano: 75.2 (#72), Llama 3.2 3B: 56.3 (#270)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Instruction Following | 1306 | 1089 |
| IFEval | 93.2% | — |
Long Context Llama 3.2 3B leads
GPT-5 Nano: 31.3 (#275), Llama 3.2 3B: 33.4 (#256)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Longer Query | 1313 | 1101 |
| Fiction.LiveBench | 44.4% | — |
Writing & Preference GPT-5 Nano leads
GPT-5 Nano: 39.4 (#244), Llama 3.2 3B: 24.9 (#302)
| Benchmark | GPT-5 Nano | Llama 3.2 3B |
|---|---|---|
| LMArena Text | 1320 | 1110 |
| LMArena Creative Writing | 1250 | 1094 |
| EQ-Bench Creative Writing | 705 | 595 |
| LMArena Multi-Turn | 1311 | 1105 |
| WildBench | 80.6% | — |
Frequently asked questions
Is GPT-5 Nano better than Llama 3.2 3B?
GPT-5 Nano is the stronger model overall, scoring 33.4 to 28.4 on the Noometry Index.
Which is cheaper, GPT-5 Nano 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; GPT-5 Nano lists at $0.05 and $0.40.
Is GPT-5 Nano or Llama 3.2 3B better for coding?
GPT-5 Nano scores higher on coding benchmarks: 33.3 versus 27.1 in the Noometry coding category.
Which has the bigger context window?
GPT-5 Nano does, with 400K tokens against 131K.
How many benchmarks do GPT-5 Nano and Llama 3.2 3B share?
15 benchmarks have published results for both models. GPT-5 Nano has 48 scored results on Noometry and Llama 3.2 3B has 18.