How much VRAM does GLM-5.3 Flash need?
GLM-5.3 Flash needs 181 GB for its weights at 4-bit, 183 GB with an 8K context and 183 GB with 32K.
- Parameters
- 321B
- Active per token
- 18B
- Longest context
- 1M
- Released
- Aug 2026
- License
- MIT
- Strongest at
- Chat, Coding, Agents, Docs
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| As released (FP8) | 306 GB | 307 GB | 307 GB | 308 GB |
| 6-bit | 245 GB | 246 GB | 247 GB | 248 GB |
| 5-bit | 213 GB | 214 GB | 214 GB | 215 GB |
| 4-bit | 181 GB | 183 GB | 183 GB | 184 GB |
| 3-bit | 146 GB | 147 GB | 148 GB | 149 GB |
| 2-bit | 112 GB | 113 GB | 114 GB | 115 GB |
| 1.6-bit | 61.7 GB | 62.8 GB | 63.1 GB | 64.3 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 | Too big | |||
| RTX 4090 | Too big | |||
| RTX 5090 | Too big | |||
| 2× RTX 3090 | Too big | |||
| RTX PRO 6000 Blackwell | Too big | |||
| M4 Max 128 GB | Too big | |||
| M5 Max 128 GB | Too big | |||
| Ryzen AI Max+ 395 128 GB | Too big | |||
| DGX Spark 128 GB | 2-bit | 104K | 26 tok/s | 3.9 s |
| M3 Ultra 256 GB | 3-bit | 1M | 43 tok/s | 11 s |
| M5 Ultra 512 GB | FP8 | 1M | 33 tok/s | 3.4 s |
| H100 SXM | Too big | |||
| DGX Station GB300 | 5-bit | 1M | 188 tok/s | 0.3 s |
| 8× H200 NVL | FP8 | 1M | 122 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
- llama.cpp
llama-server -hf unsloth/GLM-5.3-Flash-GGUF- vLLM
vllm serve zai-org/GLM-5.3-Flash- MLX (Mac)
mlx_lm.chat --model mlx-community/GLM-5.3-Flash-4bit
Questions
How much VRAM does GLM-5.3 Flash need?
GLM-5.3 Flash needs 181 GB for its weights at 4-bit, 183 GB with an 8K context and 183 GB with 32K.
Can GLM-5.3 Flash run on an RTX 4090?
Not on one card. The smallest machine here that runs it is the DGX Spark 128 GB.