4× RTX PRO 6000
81.7tok/sEveryday
1 request, prose prompt, with MTP Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality
Local AI · Recipe · 4× DGX Spark
GLM-5.2 753B INT4/INT8 with vLLM on 4× DGX Spark
GLM-5.2 753B INT4/INT8 with vLLM on 4× DGX Spark: 32.5 tok/s according to github.com (dataset as of Sep 29, 2026).
by Weschera
Intelligence (original model) · Artificial Analysisno independent value
Engine
- Engine
- vLLM
- Quantization
- QuantTrio Int4-Int8Mix (W4A16 MoE-Experts / W8A16 dense+MTP, compressed-tensors, data-free; Layer 0 + Attention-Indexer BF16); KV nvfp4_ds_mla (400 B/Token)
- Model family
- GLM-5.2
- Context
- 316,000
- Parameters
- 753B MoE, alle 256 Experts intakt (unpruned, README)
- Creator
- Weschera
- GitHub stars
- 1
- Repo updated
- Aug 20, 2026
Measurements
Every sourced value of this recipe, each with its condition and source. Bars relative to the largest value in the group.
Decode · 1 request
What you need
- Hardware
- 4 × NVIDIA DGX Spark (GB10)
- Weights
- Main weightsQuantTrio/GLM-5.2-Int4-Int8Mix MIT
- Engine
- vLLM
- Context
- 316,000 tokens
Notes
What matters before you rebuild it.
- Experimental
- Custom kernel required
Sources
Related recipes
4× RTX PRO 6000
70.7tok/sEveryday
1 request, prose prompt, with MTP Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality
4× RTX PRO 6000
64tok/sEveryday
1 request, prose prompt, with MTP (per recipe) Source
Intelligence34with thinkingno thinking 22
- ≤ 3 bit: quantization may cost quality
- Custom kernel required
4× DGX Spark
20.2tok/sEveryday
1 request, realistic prompt, with MTP (fixed-4, greedy Draft) (per recipe) Source
Intelligence34with thinkingno thinking 22
- Custom kernel required
4× DGX Spark
23.7tok/sMixed
1 request, mixed prompt set, with MTP (per recipe) Source
Intelligence34with thinkingno thinking 22
- Experimental
- Custom kernel required
3× DGX Spark
13.6tok/sMixed
1 request, mixed prompt set, with MTP (per recipe) Source
Intelligenceno independent value
- Custom kernel required
Local AI in your company?
In a workshop we work out which models and which hardware fit your tasks, and build the first agent on your infrastructure.