How much VRAM does GLM-4.7 Flash need?
GLM-4.7 Flash needs 17.6 GB for its weights at 4-bit, 18.8 GB with an 8K context and 20.1 GB with 32K. At 8-bit, 32.1 GB.
- Parameters
- 31.2B
- Active per token
- 3B
- Longest context
- 198K
- Released
- Jan 2026
- License
- MIT
- Strongest at
- Chat, Coding, Agents
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| 16-bit | 58.2 GB | 59.4 GB | 60.6 GB | 65.6 GB |
| 8-bit | 30.9 GB | 32.1 GB | 33.4 GB | 38.3 GB |
| 6-bit | 23.8 GB | 25.1 GB | 26.3 GB | 31.3 GB |
| 5-bit | 20.7 GB | 21.9 GB | 23.1 GB | 28.1 GB |
| 4-bit | 17.6 GB | 18.8 GB | 20.1 GB | 25.0 GB |
| 3-bit | 14.2 GB | 15.4 GB | 16.7 GB | 21.6 GB |
| 2-bit | 10.9 GB | 12.1 GB | 13.4 GB | 18.3 GB |
| 1.6-bit | 6.0 GB | 7.2 GB | 8.5 GB | 13.4 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | Too big | |||
| RTX 4060 Ti 16 GB | Too big | |||
| RTX 3090 | 5-bit | 24K | 165 tok/s | 0.5 s |
| RTX 4090 | 5-bit | 24K | 171 tok/s | 0.2 s |
| RTX 5090 | 6-bit | 104K | 209 tok/s | 0.2 s |
| 2× RTX 3090 | 8-bit | 198K | 130 tok/s | 0.3 s |
| RTX PRO 6000 Blackwell | 8-bit | 198K | 187 tok/s | 0.1 s |
| M4 Max 128 GB | 8-bit | 198K | 64 tok/s | 2.9 s |
| M5 Max 128 GB | 8-bit | 198K | 69 tok/s | 0.8 s |
| Ryzen AI Max+ 395 128 GB | 8-bit | 198K | 40 tok/s | 1.7 s |
| DGX Spark 128 GB | 8-bit | 198K | 51 tok/s | 0.7 s |
| M3 Ultra 256 GB | 8-bit | 198K | 80 tok/s | 1.6 s |
| M5 Ultra 512 GB | 8-bit | 198K | 94 tok/s | 0.5 s |
| H100 SXM | 8-bit | 198K | 239 tok/s | 0.1 s |
| DGX Station GB300 | 8-bit | 198K | 289 tok/s | 0.1 s |
| 8× H200 NVL | 8-bit | 198K | 265 tok/s | 0.1 s |
Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.
How to run it
- Ollama
ollama run glm-4.7-flash:latest- llama.cpp
llama-server -hf unsloth/GLM-4.7-Flash-GGUF- vLLM
vllm serve zai-org/GLM-4.7-Flash- MLX (Mac)
mlx_lm.chat --model mlx-community/GLM-4.7-Flash-4bit
Questions
How much VRAM does GLM-4.7 Flash need?
GLM-4.7 Flash needs 17.6 GB for its weights at 4-bit, 18.8 GB with an 8K context and 20.1 GB with 32K. At 8-bit, 32.1 GB.
Can GLM-4.7 Flash run on an RTX 4090?
Yes. At 5-bit it holds 24K of context and writes about 171 tokens a second.