Can glm 4 9b run on the A100 40GB?
Yes. At 8K context the best fit is F16, using about 23.2 GB of the 38.0 GB usable on the A100 40GB. Estimated speed ceiling: ~83 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | Max tok/s (est., 8K) |
|---|---|---|---|
| 2-bit (Q2_K) • | 6.9 GB Fits | 9.4 GB Fits | 390 |
| 3-bit (Q3_K_M) • | 7.9 GB Fits | 10.4 GB Fits | 307 |
| 4-bit (Q4_K_M) • | 9.0 GB Fits | 11.5 GB Fits | 249 |
| 5-bit (Q5_K_M) • | 9.8 GB Fits | 12.3 GB Fits | 218 |
| 6-bit (Q6_K) • | 10.9 GB Fits | 13.4 GB Fits | 188 |
| 8-bit (Q8_0) • | 12.5 GB Fits | 15.0 GB Fits | 156 |
| 16-bit (F16/BF16) • | 20.7 GB Fits | 23.2 GB Fits | 83 |
• real GGUF file size from mradermacher/glm-4-9b-GGUF. Speed is a bandwidth ceiling (1555 GB/s), not a benchmark. Methodology.
More
glm 4 9b on every GPU and Mac · Every model that runs on the A100 40GB · Other models that fit 48 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (zai-org/glm-4-9b), NVIDIA specs. Data as of 2026-11-17.