Can I run GLM-5.3-GGUF on an NVIDIA B300?
Yes, just
Its UD-Q2_K_XL file (254 GB) fits in the 259 GB an NVIDIA B300 can use, with room for ≈ 58K tokens of context. That file is under 3 bits per weight, so expect noticeably lower quality.
Full analysis of GLM-5.3-GGUF
Everything that runs on NVIDIA B300
Get the UD-Q2_K_XL file on Hugging Face
llama-server -hf unsloth/GLM-5.3-GGUF:UD-Q2_K_XLFiles and context on NVIDIA B300
Each file of GLM-5.3-GGUF (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) |
|---|---|---|---|---|
| BF16 | 1.51 TB | Doesn't fit | — | — |
| Q8_0 | 801 GB | Doesn't fit | — | — |
| UD-Q6_K_XL | 684 GB | Doesn't fit | — | — |
| UD-Q5_K_XL | 562 GB | Doesn't fit | — | — |
| UD-Q4_K_XL | 467 GB | Doesn't fit | — | — |
| UD-IQ4_XS | 365 GB | Doesn't fit | — | — |
| UD-Q3_K_XL | 343 GB | Doesn't fit | — | — |
| UD-IQ3_XXS | 282 GB | Doesn't fit | — | — |
| UD-Q2_K_XL | 254 GB | Tight | 58K | 116K |
| UD-IQ2_M | 239 GB | Tight | 224K | 448K |
| UD-IQ1_M | 228 GB | Tight | 334K | 667K |
| UD-IQ1_S | 217 GB | Tight | 462K | 923K |