Runtime recipe
8 × H200 SXM
141 GB per GPU
FP8 weights
vLLM model-specific guide
Single-node TP8 path. Long-context limits depend on remaining cache memory; use the guide’s matching parser and MTP settings.
GLM / MoE
GLM-5
A large sparse-attention foundation for the GLM line
New to model sizes or GPU memory? Start with weights, parameters and quantization →
GLM-5 is the useful historical baseline for understanding the 5.1–5.3 sequence. Its scale places it in a multi-GPU service class; sparse attention should be evaluated as a request-processing optimization, not a way to make the weights small.
744Breported total parameters
40Bactive parameters per token
202,752configured token positions
Publisher-reported, rounded model count; not an exact tensor census. Multimodal components and packaging can change the artifact size.
Evaluate it for
Large text-agent baselines and reproducing early GLM-5 results.
Choose another path when
A small fleet whose main constraint is resident model memory.
Release context
Why this release matters
The release emphasizes longer engineering and agent tasks. Later checkpoints refine training and attention behavior, which makes the family a good study of improvement without treating every decimal version as a new size class.
This is the dated Z.ai release-note event; a separate first weight-upload day has not been certified.
Deployment starting points
Which hardware, how many, at what precision?
Start with an exact artifact and an engine that supports it. A configuration below is evidence of a documented path; its device count does not promise a particular throughput or concurrency.
1,507.73 GB of tensor data in the pinned index
This is the index’s reported tensor payload, in decimal GB. It excludes file headers and runtime memory. Mixed precision, conversion and host/device placement determine how much GPU memory the loaded model needs.
Inspect the exact index metadata →Capacity screen: would the weights fit?
This estimates inference memory for a hypothetical uniform precision. It is useful for rejecting an allocation that is too small. It does not establish a working deployment or estimate training memory.
| Representation | Approximate weight floor | What remains to budget |
|---|---|---|
| 16-bit uniform | 1,488 GB | Cache, modality state, workspace, quantization metadata and uneven sharding |
| 8-bit uniform | 744 GB | Cache, modality state, workspace, quantization metadata and uneven sharding |
| 4-bit uniform | 372 GB | Cache, modality state, workspace, quantization metadata and uneven sharding |
Enter the total across the whole model replica, not per GPU: KV or recurrent cache for all concurrent requests, image/audio state and execution workspace. There is no universal default. We additionally leave 15% of each GPU unused.
1,488 GB for hypothetical weights
Enter a request-and-buffer budget to calculate the aggregate capacity requirement.
A count of one does not prove the engine supports this GPU. A count above one does not prove the model shards evenly or communicates efficiently. Publisher-reported, rounded model count; not an exact tensor census. Multimodal components and packaging can change the artifact size.
Architecture in practice
What the serving engine has to do
Store the history. Read the selected parts.
Fewer bytes read can reduce bandwidth pressure.
That does not prove an equal reduction in stored bytes.
Mechanism schematic based on the pinned model card and configuration. It explains a design principle; it is not a full implementation graph.
Inspect the full checkpoint specifications
| Exact checkpoint | zai-org/GLM-5 |
|---|---|
| Text model type | glm_moe_dsa |
| Layers | 78 |
| Hidden width | 6,144 |
| Attention / KV heads | 64 / 64 |
| Head dimension | 64 |
| Routed / selected experts | 256 / 8 |
| Configured positions | 202,752 |
| Documented extension | Not recorded for this checkpoint |
| Inputs → output | text → text |
| License metadata | mit |
Open weights do not imply unrestricted use. Read the applicable license terms linked from the pinned card.
A useful comparison
What else belongs on the shortlist?
5.1 is the direct generation comparison; V4 Flash provides a smaller-total sparse-attention alternative.
7 Apr 2026
GLM-5.1
Longer agent runs need process-level evaluation
744B reported total · text
24 Apr 2026
DeepSeek-V4-Flash
The smaller V4 preview that made cache design central
284B reported total · text
These are editorial comparison candidates. We have not run a matched quality or serving benchmark, so this is not a ranking.
Prepare to explain it
The interview lesson
Separate expert routing from sparse attention: they save different work and introduce different communication patterns.
Attention layouts · Quantization · Courses and worked examples
For a deployment evaluation, record exact weights, precision, engine, device count, interconnect, prompt/output lengths and concurrency. Report task success, errors, TTFT, TPOT and useful throughput together.
Sources and citation
Our assessment is editorial judgment based on the linked architecture and deployment evidence. Model facts are publisher-reported or attributed to runtime maintainers; no independent GPU benchmark was run.
AI Infra Interviews, “GLM-5: release, architecture and deployment”, checked 2026-09-13. Preserve this date and the exact checkpoint when citing.