GLM-5 · released 2026-06 · 753B parameters (40.0B active per token)
GLM-5 geometry with 1M context; only 21 of 78 layers run their own DSA indexer ('full'), the rest reuse it ('shared'). Active assumed equal to GLM-5 (40B; not restated on card).
Source configMemory for one conversation's context, on top of the weights.
| Context | FP16 | FP8 |
|---|---|---|
| 4,096 tokens | 0.34 GB | 0.17 GB |
| 8,192 tokens | 0.69 GB | 0.34 GB |
| 32,768 tokens | 2.74 GB | 1.37 GB |
| 131,072 tokens | 11.0 GB | 5.48 GB |
| 1,048,576 tokens | 87.8 GB | 43.9 GB |
| Format | Bits | Weights |
|---|---|---|
| BF16 / FP16 | 16 | 1,403.2 GB |
| FP8 (W8A8) | 8 | 705.1 GB |
| INT8 (W8A8 / W8A16) | 8 | 705.1 GB |
| NVFP4 | 4.5 | 399.7 GB |
| INT4 AWQ / GPTQ | 4.25 | 377.9 GB |
| GGUF Q8_0 | 8.5 | 745.4 GB |
| GGUF Q6_K | 6.56 | 575.3 GB |
| GGUF Q5_K_M | 5.69 | 499.0 GB |
| GGUF Q4_K_M | 4.89 | 428.8 GB |
| GGUF Q3_K_M | 3.91 | 342.9 GB |
| GGUF Q2_K | 3.35 | 293.8 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) | 8× · 70.4 tok/s | 8× · 33.9 tok/s |
| H200 SXM 141GBFP8 (W8A8) | 8× · 65.6 tok/s | 8× · 24.6 tok/s |
| H100 SXM 80GBFP8 (W8A8) | 16× · 16.7 tok/s | 16× · 11.9 tok/s |
| Instinct MI300X 192GBFP8 (W8A8) | 8× · 66.7 tok/s | 8× · 26.3 tok/s |
| A100 SXM 80GBINT4 AWQ / GPTQ | 8× · 62.2 tok/s | 8× · 18.9 tok/s |
| L40S 48GBFP8 (W8A8) | > 16 | > 16 |
| RTX PRO 6000 Blackwell Server Edition 96GBFP8 (W8A8) | 16× · 16.1 tok/s | 16× · 9.2 tok/s |
| GeForce RTX 5090 32GBGGUF Q4_K_M | > 16 | > 16 |
| GeForce RTX 4090 24GBGGUF Q4_K_M | > 16 | > 16 |
| Apple M4 Max (40-core GPU, up to 128GB)GGUF Q4_K_M | > 16 | > 16 |
| DGX Spark (GB10, 128GB unified)FP8 (W8A8) | > 16 | > 16 |