Model comparison
GPT-5 vs Llama 3.2 3B
GPT-5 is the stronger model overall, scoring 50.9 to 28.9 on the Noometry Index. Llama 3.2 3B costs 29× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Last verified . 15 shared benchmarks.
MetaSummary
- They share 15 benchmarks with published results for both. GPT-5 scores higher in 9 categories and Llama 3.2 3B in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where GPT-5 leads 63.4 to 24.7.
- The biggest single-benchmark swing is BALROG: 32.8% for GPT-5 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 $1.25 / $10 for GPT-5.
- GPT-5 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 | Llama 3.2 3B | |
|---|---|---|
| Provider | OpenAI | Meta |
| Noometry Index | 50.9 | 28.9 |
| Released | 2025-08-07 | 2024-09-24 |
| Weights | Proprietary | Open |
| Context window | 400K | 131K |
| Max output | 128K | 118K |
| Input $ / M tokens | $1.25 | $0.05 |
| Output $ / M tokens | $10 | $0.33 |
| Results tracked | 69 | 18 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 leads
GPT-5: 50.3 (#47), Llama 3.2 3B: 27.6 (#319)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Coding | 1436 | 1098 |
| SWE-bench Verified | 73.6% | — |
| SWE-bench Verified (bash only) | 65% | — |
| Aider Polyglot | 88% | — |
| LMArena WebDev | 1418 | — |
| SciCode | 42.9% | — |
| GSO | 6.9% | — |
| WeirdML | 60.7% | — |
| BigCodeBench Instruct | — | 23.4% |
| BigCodeBench Complete | — | 28.3% |
| ALE-Bench | 1,162 | — |
| AlgoTune | 1.67 | — |
Agentic & Tool Use GPT-5 leads
GPT-5: 33.1 (#56), Llama 3.2 3B: 20.1 (#143)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| BALROG | 32.8% | 10.1% |
| Terminal-Bench | 49.6% | — |
| Berkeley Function Calling Leaderboard | — | 21.9% |
| GDPval | 34.8% | — |
| Remote Labor Index | 1.7% | — |
| DeepResearch Bench | 49.6% | — |
| LMArena Search | 1133 | — |
| METR Time Horizons | 69.6% | — |
Reasoning GPT-5 leads
GPT-5: 38.3 (#64), Llama 3.2 3B: 21.0 (#228)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Hard Prompts | 1416 | 1095 |
| ARC-AGI-2 | 9.9% | — |
| SimpleBench | 56.7% | — |
| Kagi LLM Benchmark | 72.7% | — |
| ARC-AGI-1 | 65.7% | — |
| CritPt | 12.6% | — |
| Chess Puzzles | 37% | — |
| EnigmaEval | 10.5% | — |
| EBR-Bench | 12.7% | — |
| Mystery Game Puzzles | 23% | — |
| DTBench | 90.7% | — |
| LMCA | 40% | — |
| Epoch Capabilities Index | 150 | — |
| ForecastBench | 61.4 | — |
Math GPT-5 leads
GPT-5: 55.0 (#44), Llama 3.2 3B: 32.4 (#214)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Math | 1407 | 1126 |
| FrontierMath (Tiers 1-3) | 55.4% | — |
| FrontierMath Tier 4 | 22% | — |
| OTIS Mock AIME 2024-2025 | 91.4% | — |
| ProofBench | 18% | — |
| Omni-MATH | 64.7% | — |
| MATH Level 5 | 98.1% | — |
| FrontierMath (Feb 2025 set) | 32.4% | — |
| FrontierMath Tier 4 (v1) | 12.5% | — |
Knowledge GPT-5 leads
GPT-5: 56.6 (#43), Llama 3.2 3B: 29.7 (#235)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Expert | 1419 | 1090 |
| GPQA Diamond | 86.2% | — |
| Humanity's Last Exam | 25.3% | — |
| SimpleQA Verified | 50.1% | — |
| MMLU-Pro | 86.3% | — |
| Confabulations | 10.3% | — |
| Vectara Hallucination Rate | 14.7% | — |
| GPQA (HELM) | 79.2% | — |
Multimodal Not comparable
GPT-5: 46.8 (#13), Llama 3.2 3B: —
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Vision | 1232 | — |
| GeoBench | 81% | — |
| VPCT | 66% | — |
Multilingual GPT-5 leads
GPT-5: 51.4 (#110), Llama 3.2 3B: 26.2 (#281)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Non-English | 1397 | 1019 |
| LMArena Chinese | 1422 | 1017 |
| LMArena German | 1416 | 1056 |
| LMArena Russian | 1406 | 949 |
| LMArena French | 1410 | — |
| LMArena Japanese | 1409 | — |
| LMArena Korean | 1360 | — |
| LMArena Spanish | 1399 | — |
Instruction Following GPT-5 leads
GPT-5: 73.8 (#113), Llama 3.2 3B: 56.0 (#275)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Instruction Following | 1388 | 1089 |
| IFEval | 87.5% | — |
Long Context GPT-5 leads
GPT-5: 69.5 (#2), Llama 3.2 3B: 33.4 (#261)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Longer Query | 1399 | 1100 |
| Fiction.LiveBench | 97.2% | — |
Writing & Preference GPT-5 leads
GPT-5: 63.4 (#65), Llama 3.2 3B: 24.7 (#307)
| Benchmark | GPT-5 | Llama 3.2 3B |
|---|---|---|
| LMArena Text | 1406 | 1110 |
| LMArena Creative Writing | 1365 | 1094 |
| EQ-Bench Creative Writing | 1627 | 595 |
| LMArena Multi-Turn | 1426 | 1105 |
| Short-Story Creative Writing | 86% | — |
| WildBench | 85.7% | — |
Frequently asked questions
Is GPT-5 better than Llama 3.2 3B?
GPT-5 is the stronger model overall, scoring 50.9 to 28.9 on the Noometry Index. Llama 3.2 3B costs 29× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Which is cheaper, GPT-5 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 lists at $1.25 and $10.
Is GPT-5 or Llama 3.2 3B better for coding?
GPT-5 scores higher on coding benchmarks: 50.3 versus 27.6 in the Noometry coding category.
Which has the bigger context window?
GPT-5 does, with 400K tokens against 131K.
How many benchmarks do GPT-5 and Llama 3.2 3B share?
15 benchmarks have published results for both models. GPT-5 has 69 scored results on Noometry and Llama 3.2 3B has 18.