flint
flint is an open-weights language model from RecursiveSI, released October 21, 2026. It has 40B parameters (mixture-of-experts, 1.3B active). The context window is 256K tokens.
Check it against your own PC →
Can you run flint locally?
The smallest common GPU it runs on is 6 GB of VRAM (with 32 GB of RAM): Q5_K_M, GPU + experts in RAM, roughly 49 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 | Runs · GPU + CPU experts | Q5_K_M (excellent) | 5.2 GB | 24.2 GB | ~49 |
| 8 GB VRAM | Runs · GPU + CPU experts | Q5_K_M (excellent) | 7.2 GB | 22.2 GB | ~54 |
| 12 GB VRAM | Runs · GPU + CPU experts | Q6_K (≈ lossless) | 11.2 GB | 22.5 GB | ~54 |
| 16 GB VRAM | Runs · GPU + CPU experts | Q6_K (≈ lossless) | 15.2 GB | 18.5 GB | ~64 |
| 24 GB VRAM | Runs on GPU | IQ4_XS (good) | 22.2 GB | – | ~460 |
| 32 GB VRAM | Runs on GPU | Q5_K_M (excellent) | 29.4 GB | – | ~550 |
Ready-made quantizations on Hugging Face: unsloth/flint-GGUF · lmstudio-community/flint-GGUF · ggml-org/Flint-GGUF · bartowski/RecursiveSI_flint-GGUF · AtomicChat/Flint-GGUF · anik-jha/Flint-40B-A2.5B-8GB-GGUF
Benchmark results
Epoch AI has not published results for flint yet. New models are usually evaluated within days to weeks.