All models › google
Arena score1443#67 of 416 in Chat, paid models included
Size31.3 Bparameters
4-bit download19 GBweights only, Q4_K_M
Max context256Ktokens, from the model config
Where to use it for free
- Free API on OpenRouter: model id
google/gemma-4-31b-it:free. Try it side by side · OpenRouter page - The lab's own chat: aistudio.google.com (free with daily limits)
- Download and run it at home: official weights · GGUF files for Ollama and LM Studio
Arena scores
Rank among all models, paid ones included. Source: Arena leaderboard dataset.
Speed on popular hardware
| Hardware | GPU memory | Fits? | Speed |
|---|---|---|---|
| RTX 3060 12 GB | 12 GB | GPU + 64 GB RAM | 4.5 tok/s |
| RTX 4060 Ti 16 GB | 16 GB | GPU + 64 GB RAM | 4.9 tok/s |
| RTX 3090 | 24 GB | GPU + 64 GB RAM | 17 tok/s |
| RTX 4090 | 24 GB | GPU + 64 GB RAM | 18 tok/s |
| RTX 5090 | 32 GB | Fits in GPU | 50 tok/s |
| RX 7900 XTX | 24 GB | GPU + 64 GB RAM | 17 tok/s |
| 2 × RTX 3090 | 48 GB | Fits in GPU | 25 tok/s |
| 4 × RTX 3090 | 96 GB | Fits in GPU | 24 tok/s |
| Mac M4 Max 128 GB | 96 GB | Fits in GPU | 15 tok/s |
| Ryzen AI Max+ 395 128 GB | 96 GB | Fits in GPU | 7.1 tok/s |
| 8 × H100 SXM | 640 GB | Fits in GPU | 77 tok/s |
Estimate at 4-bit with 32K context, 64 GB of DDR5 for offload. Speed ≈ memory bandwidth × 70% ÷ data read per token; real numbers vary by about ±30%.
GPU memory needed
| Level | 8K context | 32K context | 128K context | Quality vs full |
|---|---|---|---|---|
| Q3_K_M | 18 GB | 22 GB | 39 GB | +6.9% noticeable loss |
| Q4_K_M | 22 GB | 26 GB | 42 GB | +1.8% small loss |
| Q5_K_M | 25 GB | 29 GB | 45 GB | +0.6% near-lossless |
| Q6_K | 28 GB | 32 GB | 49 GB | +0.2% near-lossless |
| Q8_0 | 36 GB | 40 GB | 56 GB | ≈ 0% same as full |
| F16 | 65 GB | 69 GB | 86 GB | ≈ 0% same as full |
Weights + context cache (from the model's architecture) + runtime. Quality: increase in perplexity against the full model, from llama.cpp tests, scaled by model size. Simulate your own setup.
Chat score over time
145015001550
MayJunJulAugSeptOct
Gemma 4 31B 1443Best paid model
Arena score from 17 Apr 2026 to 15 Oct 2026.
Vision score over time
12501275130013251350
MayJunJulAugSeptOct
Gemma 4 31B 1277Best paid model
Arena score from 10 Apr 2026 to 9 Oct 2026.
Similar free models in Chat
- kimi-k2.5-thinking 1445
- deepseek-v4-pro-high-preview 1445
- nvidia-nemotron-3-ultra-550b-a55b-nvfp4 1445 · 560.5 B
- Inkling 1441
- Qwen3.8-27B 1441 · 27.8 B
- GLM-5 1446 · 753.9 B