INTELLECT-3
INTELLECT-3 is an open-weights language model from Prime Intellect, released November 26, 2025. It has 110B parameters (mixture-of-experts, 17B active). The context window is 128K tokens.
We present INTELLECT-3, a 106B-parameter Mixture-of-Experts model (12B active) trained with large-scale reinforcement learning on our end-to-end RL infrastructure stack.
Check it against your own PC →
Can you run INTELLECT-3 locally?
The smallest common GPU it runs on is 8 GB of VRAM (with 32 GB of RAM): IQ2_XXS, GPU + experts in RAM, roughly 14 tokens per second. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Too large | – | – | – | – |
| 8 GB VRAM | Runs · GPU + CPU experts | IQ2_XXS (heavy loss) | 7.2 GB | 24.9 GB | ~14 |
| 12 GB VRAM | Runs · GPU + CPU experts | IQ2_XXS (heavy loss) | 11.2 GB | 20.9 GB | ~16 |
| 16 GB VRAM | Runs · GPU + CPU experts | IQ2_M (heavy loss) | 15.2 GB | 25.8 GB | ~14 |
| 24 GB VRAM | Runs · GPU + CPU experts | IQ3_XXS (noticeable loss) | 23.2 GB | 22.7 GB | ~17 |
| 32 GB VRAM | Runs · GPU + CPU experts | IQ3_XXS (noticeable loss) | 31.2 GB | 14.7 GB | ~26 |
Benchmark results
Epoch AI has not published results for INTELLECT-3 yet. New models are usually evaluated within days to weeks.
Links
Announcement · Hugging Face: PrimeIntellect/INTELLECT-3-Base