Can I run GLM-5.3-Flash on an NVIDIA B300?
Yes, with a GGUF version
The published weights are too big, but the UD-Q4_K_XL file from unsloth/GLM-5.3-Flash-GGUF (200 GB) fits, with room for an unknown amount of context.
Full analysis of GLM-5.3-Flash
Everything that runs on NVIDIA B300
Get the UD-Q4_K_XL file on Hugging Face
llama-server -hf unsloth/GLM-5.3-Flash-GGUF:UD-Q4_K_XLFiles and context on NVIDIA B300
Each file of GLM-5.3-Flash (and of GGUF versions the site has analyzed), and how much context fits next to it, with the context cache in 16-bit or 8-bit.
| File | Size | On NVIDIA B300 | Context (16-bit) | Context (8-bit) |
|---|---|---|---|---|
| As published | 328 GB | Doesn't fit | — | — |
| BF16 | 642 GB | Doesn't fit | — | — |
| Q8_0 | 341 GB | Doesn't fit | — | — |
| UD-Q6_K_XL | 292 GB | Doesn't fit | — | — |
| UD-Q5_K_XL | 240 GB | Doesn't fit | — | — |
| UD-Q4_K_XL | 200 GB | Fits | not known | not known |
| UD-IQ4_XS | 157 GB | Fits | not known | not known |
| UD-Q3_K_XL | 148 GB | Fits | not known | not known |
| UD-IQ3_XXS | 120 GB | Fits | not known | not known |
| UD-Q2_K_XL | 109 GB | Fits | not known | not known |
| UD-IQ2_XXS | 102 GB | Fits | not known | not known |
| UD-IQ1_M | 97.6 GB | Fits | not known | not known |
| UD-IQ1_S | 93.1 GB | Fits | not known | not known |