Model comparison
Gemma 3 4B vs GPT-5 Nano
GPT-5 Nano is the stronger model overall, scoring 33.5 to 28.1 on the Noometry Index. Gemma 3 4B costs 2.8× less per token, which makes it the better buy when GPT-5 Nano's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
GoogleSummary
- They share 22 benchmarks with published results for both. Gemma 3 4B scores higher in 3 categories and GPT-5 Nano in 6 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where GPT-5 Nano leads 35.9 to 11.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 7.5% for Gemma 3 4B and 81.1% for GPT-5 Nano.
- Gemma 3 4B is cheaper at $0.04 / $0.08 per million input/output tokens, against $0.05 / $0.40 for GPT-5 Nano.
- GPT-5 Nano accepts more context: 400K tokens versus 131K.
- Gemma 3 4B has downloadable open weights; the other is API-only.
Side by side
| Gemma 3 4B | GPT-5 Nano | |
|---|---|---|
| Provider | OpenAI | |
| Noometry Index | 28.1 | 33.5 |
| Released | 2025-03-12 | 2025-08-07 |
| Weights | Open | Proprietary |
| Context window | 131K | 400K |
| Max output | 4K | 128K |
| Input $ / M tokens | $0.04 | $0.05 |
| Output $ / M tokens | $0.08 | $0.40 |
| Results tracked | 22 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemma 3 4B leads
Gemma 3 4B: 35.9 (#215), GPT-5 Nano: 33.6 (#254)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| LMArena Coding | 1230 | 1351 |
| SWE-bench Verified (bash only) | — | 34.8% |
| WeirdML | — | 38.1% |
| ALE-Bench | — | 718.67 |
Agentic & Tool Use GPT-5 Nano leads
Gemma 3 4B: 20.9 (#142), GPT-5 Nano: 25.8 (#106)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| Berkeley Function Calling Leaderboard | 19.6% | 51.5% |
| Terminal-Bench | — | 21.8% |
Reasoning GPT-5 Nano leads
Gemma 3 4B: 13.2 (#335), GPT-5 Nano: 16.3 (#306)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| Kagi LLM Benchmark | 25.2% | 62.2% |
| Chess Puzzles | 0% | 27% |
| LMArena Hard Prompts | 1253 | 1328 |
| DTBench | 50.9% | 62.7% |
| LMCA | 2.8% | 7.9% |
| Epoch Capabilities Index | 116.02 | 139.38 |
| ARC-AGI-2 | — | 2.6% |
| ARC-AGI-1 | — | 20.7% |
| Mystery Game Puzzles | — | 9% |
| ForecastBench | — | 59.1 |
Math GPT-5 Nano leads
Gemma 3 4B: 16.8 (#292), GPT-5 Nano: 29.4 (#241)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 7.5% | 81.1% |
| LMArena Math | 1239 | 1317 |
| FrontierMath (Tiers 1-3) | — | 20% |
| FrontierMath Tier 4 | — | 2.4% |
| 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
Gemma 3 4B: 11.8 (#299), GPT-5 Nano: 35.9 (#178)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| GPQA Diamond | 23.2% | 69.4% |
| Vectara Hallucination Rate | 6.4% | 10.5% |
| LMArena Expert | 1223 | 1321 |
| SimpleQA Verified | — | 11.7% |
| MMLU-Pro | — | 77.8% |
| GPQA (HELM) | — | 67.9% |
Multimodal Not comparable
Gemma 3 4B: —, GPT-5 Nano: 31.3 (#108)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| LMArena Vision | — | 1159 |
| VPCT | — | 37.2% |
Multilingual GPT-5 Nano leads
Gemma 3 4B: 42.5 (#194), GPT-5 Nano: 45.3 (#172)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| LMArena Non-English | 1273 | 1313 |
| LMArena German | 1281 | 1327 |
| LMArena Russian | 1294 | 1296 |
| LMArena Chinese | — | 1356 |
| LMArena Japanese | — | 1226 |
| LMArena Korean | — | 1269 |
| LMArena Spanish | — | 1360 |
Instruction Following GPT-5 Nano leads
Gemma 3 4B: 65.2 (#225), GPT-5 Nano: 75.0 (#79)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| LMArena Instruction Following | 1239 | 1306 |
| IFEval | — | 93.2% |
Long Context Gemma 3 4B leads
Gemma 3 4B: 38.7 (#194), GPT-5 Nano: 31.3 (#281)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| LMArena Longer Query | 1273 | 1312 |
| Fiction.LiveBench | — | 44.4% |
Writing & Preference Gemma 3 4B leads
Gemma 3 4B: 42.0 (#239), GPT-5 Nano: 39.1 (#249)
| Benchmark | Gemma 3 4B | GPT-5 Nano |
|---|---|---|
| LMArena Text | 1291 | 1320 |
| LMArena Creative Writing | 1271 | 1249 |
| EQ-Bench Creative Writing | 1068 | 705 |
| LMArena Multi-Turn | 1255 | 1311 |
| WildBench | — | 80.6% |
Frequently asked questions
Is Gemma 3 4B better than GPT-5 Nano?
GPT-5 Nano is the stronger model overall, scoring 33.5 to 28.1 on the Noometry Index. Gemma 3 4B costs 2.8× less per token, which makes it the better buy when GPT-5 Nano's lead doesn't matter for your workload.
Which is cheaper, Gemma 3 4B or GPT-5 Nano?
Gemma 3 4B is cheaper. It lists at $0.04 per million input tokens and $0.08 per million output tokens; GPT-5 Nano lists at $0.05 and $0.40.
Is Gemma 3 4B or GPT-5 Nano better for coding?
Gemma 3 4B scores higher on coding benchmarks: 35.9 versus 33.6 in the Noometry coding category.
Which has the bigger context window?
GPT-5 Nano does, with 400K tokens against 131K.
How many benchmarks do Gemma 3 4B and GPT-5 Nano share?
22 benchmarks have published results for both models. Gemma 3 4B has 22 scored results on Noometry and GPT-5 Nano has 49.