GLM-4.5 · released 2026-01 · 31.2B parameters (3.90B active per token)
MLA (Glm4MoeLite); MLA cache 576 elements/token/layer. params incl. 1 MTP layer.
Source configMemory for one conversation's context, on top of the weights.
| Context | FP16 | FP8 |
|---|---|---|
| 4,096 tokens | 0.21 GB | 0.10 GB |
| 8,192 tokens | 0.41 GB | 0.21 GB |
| 32,768 tokens | 1.65 GB | 0.83 GB |
| 131,072 tokens | 6.61 GB | 3.30 GB |
| 202,752 tokens | 10.2 GB | 5.11 GB |
| Format | Bits | Weights |
|---|---|---|
| BF16 / FP16 | 16 | 58.2 GB |
| FP8 (W8A8) | 8 | 30.3 GB |
| INT8 (W8A8 / W8A16) | 8 | 30.3 GB |
| NVFP4 | 4.5 | 18.1 GB |
| INT4 AWQ / GPTQ | 4.25 | 17.2 GB |
| GGUF Q8_0 | 8.5 | 30.9 GB |
| GGUF Q6_K | 6.56 | 23.8 GB |
| GGUF Q5_K_M | 5.69 | 20.7 GB |
| GGUF Q4_K_M | 4.89 | 17.8 GB |
| GGUF Q3_K_M | 3.91 | 14.2 GB |
| GGUF Q2_K | 3.35 | 12.2 GB |
One server, 8K context, FP8 on data center GPUs and GGUF Q4_K_M on consumer and Apple hardware.
| GPU | 1 user | 32 users |
|---|---|---|
| B200 SXM (HGX B200)FP8 (W8A8) | 1× · 379 tok/s | 1× · 105 tok/s |
| H200 SXM 141GBFP8 (W8A8) | 1× · 287 tok/s | 1× · 66.8 tok/s |
| H100 SXM 80GBFP8 (W8A8) | 1× · 227 tok/s | 1× · 47.9 tok/s |
| Instinct MI300X 192GBFP8 (W8A8) | 1× · 304 tok/s | 1× · 73.1 tok/s |
| A100 SXM 80GBINT4 AWQ / GPTQ | 1× · 233 tok/s | 1× · 44.9 tok/s |
| L40S 48GBFP8 (W8A8) | 1× · 76.0 tok/s | 1× · 13.0 tok/s |
| RTX PRO 6000 Blackwell Server Edition 96GBFP8 (W8A8) | 1× · 129 tok/s | 1× · 23.7 tok/s |
| GeForce RTX 5090 32GBGGUF Q4_K_M | 1× · 149 tok/s | 1× · 33.1 tok/s |
| GeForce RTX 4090 24GBGGUF Q4_K_M | 1× · 104 tok/s | 2× · 26.6 tok/s |
| Apple M4 Max (40-core GPU, up to 128GB)GGUF Q4_K_M | 1× · 75.3 tok/s | 1× · 12.9 tok/s |
| DGX Spark (GB10, 128GB unified)FP8 (W8A8) | 1× · 25.8 tok/s | 1× · 4.2 tok/s |