unsloth
GLM 4.7 Flash
GLM 4.7 Flash is a 29.9B mixture-of-experts deepseek2 model from unsloth, with about 6.4B parameters active per token. Configured context length is 202,752 tokens. License: mit. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters29.9B / 6.4B active
Context198K
Attentionmla
Licensemit
47 layers · 20 KV heads · head dim 102 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 55.8 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 33.2 GB | Hugging Face file bytes |
| Q8_0 | 29.7 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 24.4 GB | Hugging Face file bytes |
| Q6_K | 23.0 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 20.2 GB | Hugging Face file bytes |
| Q5_K_M | 19.9 GB | Hugging Face file bytes |
| Q5_K_S | 19.4 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 16.3 GB | Hugging Face file bytes |
| Q4_K_M | 17.1 GB | Hugging Face file bytes |
| Q4_K_S | 16.1 GB | Hugging Face file bytes |
| IQ4_NL | 16.0 GB | Hugging Face file bytes |
| IQ4_XS | 15.2 GB | Hugging Face file bytes |
| Q4_1 | 17.7 GB | Hugging Face file bytes |
| Q4_0 | 16.0 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 12.8 GB | Hugging Face file bytes |
| Q3_K_M | 13.6 GB | Hugging Face file bytes |
| Q3_K_S | 12.4 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 12.0 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 11.1 GB | Hugging Face file bytes |
| Q2_K | 21.2 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 9.79 GB | Hugging Face file bytes |
| UD-IQ1_M | 9.13 GB | Hugging Face file bytes |
| UD-IQ1_S | 8.61 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Expert offload | UD-Q3_K_XL | 47–64 |
| GeForce RTX 3060 12GB | Tight fit | UD-IQ2_XXS | 74–101 |
| GeForce RTX 4060 Ti 16GB | Tight fit | UD-Q3_K_XL | 47–64 |
| GeForce RTX 4090 | Tight fit | UD-Q5_K_XL | 110–150 |
| GeForce RTX 5090 | Runs | UD-Q6_K_XL | 164–224 |
| Apple M4 Pro | Tight fit | UD-Q4_K_XL | 36–49 |
| Apple M5 | Tight fit | UD-IQ2_XXS | 31–43 |
| CPU only · DDR5-5600 | Runs on CPU | Q6_K | 8.7–12 |
Smallest devices that fit it fully
- GeForce RTX 2080 Ti — Tight fit, UD-IQ1_M, 147–200 tok/s
- GeForce GTX 1080 Ti — Tight fit, UD-IQ1_M, 115–157 tok/s
- GeForce RTX 5070 — Tight fit, UD-IQ2_XXS, 138–188 tok/s
- GeForce RTX 4070 Ti — Tight fit, UD-IQ2_XXS, 104–141 tok/s
- GeForce RTX 4070 Super — Tight fit, UD-IQ2_XXS, 104–141 tok/s
- GeForce RTX 4070 — Tight fit, UD-IQ2_XXS, 104–141 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-Q3_K_XL).
llama.cpp
llama-server -hf unsloth/GLM-4.7-Flash-GGUF:UD-Q3_K_XL -c 8192 -ngl 47
Ollama
ollama run glm-4.7-flash
LM Studio
lms get unsloth/GLM-4.7-Flash-GGUF # Then load quant UD-Q3_K_XL