GLM-5 models
23 GLM-5 models from Z.ai on Hugging Face, from 1.2B to 753.9B parameters, published by Z.ai in F8_E4M3 and BF16, with contexts from 198K to 1M tokens. The smallest official model, GLM-5.3-Flash, needs about 359 GB of VRAM at its published precision; the cheapest live fit is RTX PRO 6000 at $5.50/hr.
Part of the GLM series · Previous generation: GLM-4.5
Pick a size
One row per official GLM-5 size: the VRAM it needs at each precision, the cheapest GPU that holds it today, and the KV cache for a 32K-token context.
| Model | Parameters | Native VRAM | FP8 VRAM | INT4 VRAM | Live GPU fit (native) | Est. $/hr | KV cache at 32K |
|---|---|---|---|---|---|---|---|
| GLM-5.3-Flash | 321.3B | 359 GB | – | 180 GB | RTX PRO 6000 × 4 | $5.50/hr | not published for this architecture |
| GLM-5.2 | 753.3B | 1684 GB | 842 GB | 421 GB | No live fit | – | 2.7 GB |
| GLM-5.1 | 753.9B | 1685 GB | 843 GB | 421 GB | No live fit | – | 2.7 GB |
VRAM is weights times a flat 1.2 overhead; the KV cache is a separate per-model figure at 16-bit, shown where the architecture is published. See the methodology.
Official models (7)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| GLM-5.3-Flash | 321.3B | F8_E4M3 | 359 GB | RTX PRO 6000 × 4 | $5.50/hr |
| GLM-5.3-Flash-BF16 | 321.3B | BF16 | 718 GB | RTX PRO 6000 × 8 | $11.00/hr |
| GLM-5.2 | 753.3B | BF16 | 1684 GB | No live fit | – |
| GLM-5.3 | 753.3B | F8_E4M3 | 842 GB | No live fit | – |
| GLM-5.3-BF16 | 753.3B | BF16 | 1684 GB | No live fit | – |
| GLM-5 | 753.9B | BF16 | 1685 GB | No live fit | – |
| GLM-5.1 | 753.9B | BF16 | 1685 GB | No live fit | – |
Quantized versions (9)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| GLM-5.3-Flash-ABLITERATED-FP8 | 321.3B | F8_E4M3 | 359 GB | RTX PRO 6000 × 4 | $5.50/hr |
| GLM-5.3-Flash-UNCENSORED-FP8 | 321.3B | F8_E4M3 | 359 GB | RTX PRO 6000 × 4 | $5.50/hr |
| GLM-5.3-Flash-Uncensored-FP8 | 321.3B | F8_E4M3 | 359 GB | RTX PRO 6000 × 4 | $5.50/hr |
| GLM-5.3-Flash-FP8 | 321.3B | F8_E4M3 | 359 GB | RTX PRO 6000 × 4 | $5.50/hr |
| GLM-5.3-CYBERSECURITY-FP8 | 753.3B | F8_E4M3 | 842 GB | No live fit | – |
| GLM-5.3-UNCENSORED-FP8 | 753.3B | F8_E4M3 | 842 GB | No live fit | – |
| GLM-5.2-FP8 | 753.3B | F8_E4M3 | 842 GB | No live fit | – |
| GLM-5-FP8 | 753.9B | F8_E4M3 | 843 GB | No live fit | – |
| GLM-5.1-FP8 | 753.9B | F8_E4M3 | 843 GB | No live fit | – |
Fine-tunes and community models (7)
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| GLM-5.3-Flash-DFlash2 | 1.2B | BF16 | 2.6 GB | RTX 3070 | $0.132/hr |
| apex-flash-1 | 321.3B | BF16 | 718 GB | RTX PRO 6000 × 8 | $11.00/hr |
| apex-flash-1-abliterated | 321.3B | BF16 | 718 GB | RTX PRO 6000 × 8 | $11.00/hr |
| GLM-5.3-Flash | 321.3B | BF16 | 718 GB | RTX PRO 6000 × 8 | $11.00/hr |
| penclaw-GLM-5.3-abliterated | 753.3B | BF16 | 1684 GB | No live fit | – |
| GLM-5.3-F.U-AnthraClaud-Edition-BF16 | 753.3B | BF16 | 1684 GB | No live fit | – |
| Hydrogen-1.0 | 753.3B | BF16 | 1684 GB | No live fit | – |
VRAM is for the precision the model is published in: the weights times a flat 1.2 overhead, with the KV cache not included. See the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.
Run and fine-tune GLM-5
- How to run GLM-5: engine support, chat template and context flags, with a serving command sized per model.
- How to fine-tune GLM-5: LoRA and QLoRA memory per model, and the license limits on derivatives.
