Compute Exchange launches a secondary GPU Marketplace
Compute Exchange
Sign Up
MODEL FAMILY · GLM

FLAGSHIP OPEN-WEIGHT
BILINGUAL REASONING.

GLM is Z.ai’s open-weight family of large language models, led by GLM-5.2 — a 753B-parameter multimodal flagship with 1M-token context, FP8 inference, native tool calling, code generation, explicit reasoning, and vision inputs. Procure GLM inference through Compute Exchange Token Forwards or Reserved GPU Rental.

FAMILY LINEUP

GLM-5 AND BEYOND

The GLM-5 line is the current flagship; GLM-4.7-Flash covers latency-sensitive workloads; GLM-OCR specializes in document and vision extraction. All ship under MIT license as open-weight releases from Z.ai.

GLM-5.2

FLAGSHIP · CURRENT

Z.ai's latest flagship multimodal model — strong bilingual (Chinese-English) reasoning, long-context understanding (up to 1M tokens), vision inputs, advanced tool use, code generation, and agent-oriented behavior.

Params
753B
Context
1M
Quant
FP8
Modality
MULTIMODAL
Tool callingReasoningCode genVisionResponses API
License: MIT

GLM-5.1

PREVIOUS FLAGSHIP

Previous-generation flagship in the GLM-5 line — strong bilingual reasoning, long-horizon agentic tasks (sustains thousands of tool calls), and SWE-Bench Pro state-of-the-art code generation.

Params
754B
Context
200K
Quant
FP8
Modality
TEXT-TO-TEXT
Tool callingReasoningCode gen
License: MIT

GLM-4.7-Flash

FAST / LIGHTWEIGHT MOE

30B-parameter MoE with 3B active per token — preserved thinking mode for multi-turn agentic tasks, with speculative decoding and multi-token prediction for low-latency, high-throughput inference.

Params
30B-A3B
Context
200K
Quant
BF16
Modality
TEXT-TO-TEXT
Tool callingReasoningLow latency
License: MIT

GLM-OCR

VISION / OCR SPECIALIST

CogViT visual encoder + GLM-0.5B language decoder for OCR, document parsing, formula and table recognition. #1 on OmniDocBench V1.5 (94.62). *PP-DocLayoutV3 sub-component under Apache 2.0.

Params
0.9B
Context
128K
Quant
—
Modality
IMAGE-TEXT-TO-TEXT
VisionOCRDoc parsingTables
License: MIT*
CAPABILITIES

WHAT GLM DOES WELL

BILINGUAL REASONING

Strong Chinese-English reasoning across long-form text, code, and structured tasks. Competitive with Western flagships on English benchmarks and class-leading on Chinese.

1M LONG CONTEXT

Up to 1M tokens on GLM-5.2 (with sparse-attention IndexShare reducing per-token FLOPs ~2.9× at full context) — absorb full legal filings, codebases, or research corpora without chunking.

NATIVE TOOL USE

First-class function calling, structured outputs, and Responses API support. Agent-oriented architecture handles multi-step plans and tool composition.

EXPLICIT REASONING

Reasoning mode surfaces chain-of-thought scratchpads for complex math, code, and analytical tasks. Tunable depth at the API boundary.

MULTIMODAL INPUTS

GLM-5.2 accepts image and text inputs natively. Pair with the specialized GLM-OCR (0.9B, CogViT + GLM-0.5B; #1 on OmniDocBench V1.5) for high-volume document extraction pipelines.

FP8 EFFICIENCY

Native FP8 quantization keeps cost-per-token competitive and inference fast on H100-class hardware across the open-weight provider network.

WHERE GLM FITS

REPRESENTATIVE WORKLOADS

BILINGUAL ENTREPRISE ASSISTANTS

Customer-facing or internal assistants serving Chinese-English markets with consistent reasoning quality across both languages.

LONG DOCUMENT ANALYSIS

Legal filings, financial disclosures, technical specifications — 432K context absorbs full documents without retrieval chunking.

AGENTIC WORKFLOWS

Tool-calling backbone for multi-step agents — research loops, code generation pipelines, structured action sequences.

RAG WITH REDUCED RETRIEVAL

Long-context tolerance lets you pack more context per query and reduce the brittleness of retrieval recall.

MULTIMODAL DOCUMENT PIPELINES

GLM-OCR for visual extraction → GLM-5.2 for reasoning over extracted content. End-to-end open-weight document understanding.

OPEN-WEIGHT PRODUCTION INFERENCE

MIT-licensed alternative to closed flagships, with deployable weights for sovereign and on-prem buyers.

PROCUREMENT

HOW TO ACCESS GLM

Two procurement paths through Compute Exchange. Choose by who you want operating the model.

Token Forwards

COMMITTED INFERENCE

Lock GLM inference capacity in advance, denominated in Standardized Token Units. Provider operates the model; you tap tokens against a committed balance over terms up to six months.

  • · Provider operates and scales GLM endpoint
  • · Per-STU rate locked at commitment
  • · Realtime, batch, or mixed latency
  • · Quotes against the published STU index
Token Forwards
Reserved GPU Rental

RUN YOUR OWN

Reserve H100-class capacity from the neocloud network and deploy GLM weights yourself. Full operational control — sovereign data, custom serving stack, fine-tuned variants.

  • · MIT-licensed weights — deploy anywhere
  • · Custom serving stack (vLLM, SGLang, TensorRT-LLM)
  • · Sovereign / on-prem / air-gapped deployments
  • · Terms from 1 month to 24+ months
Reserved GPU Rental
Frequently Asked Questions

GLM, EXPLAINED

Who builds GLM?+
GLM (General Language Model) is developed by Z.ai (formerly Zhipu AI), a Beijing-based AI company spun out of Tsinghua University. Z.ai develops the ChatGLM family of open-weight models and contributes to LLMs-as-Agents research. Weights are publicly released under the zai-org namespace.
What is the difference between GLM-5.2 and GLM-5.1?+
Both are 753B+ parameter flagships in the GLM-5 line under MIT license, with native FP8 quantization and tool calling and reasoning support. GLM-5.2 is multimodal (accepts image and text inputs), adds a 1M-token context window via the IndexShare sparse-attention architecture (~2.9× FLOPs reduction at full context), Responses API support, and incremental gains on code and agentic benchmarks. GLM-5.1 is text-to-text and holds state-of-the-art SWE-Bench Pro (58.4) — well-suited for long-horizon agentic workloads at slightly lower cost. For high-volume OCR / document parsing, pair either with the specialized GLM-OCR model.
Is GLM open-weight?+
Yes. The full GLM-5 family is released under the MIT license, including FP8 quantized variants for efficient deployment. This makes GLM suitable for sovereign-data, on-prem, and air-gapped deployments where buyers need to operate the model themselves rather than rent inference from a third party.
How does GLM compare to Western flagship models?+
GLM-5.2 is competitive on bilingual reasoning, tool use, and long-context tasks. It particularly outperforms Western models on Chinese-language workloads given training-data emphasis, and matches them on many English benchmarks. Specific benchmark parity varies by task — request a quote with your workload context and we surface representative benchmarks per provider.
How do I procure GLM inference through Compute Exchange?+
GLM is available through Token Forwards (committed-use inference capacity in Standardized Token Units) and Reserved GPU Rental (run your own GLM deployment on dedicated GPUs). Submit an RFQ specifying GLM-5.2 or GLM-5.1, your commitment size, and your term — Compute Exchange returns quotes from the verified open-model provider network.
How does GLM map to the STU methodology?+
GLM-5.2's native output-to-input ratio is approximately 3.14× (one output token costs about 3.14× one uncached input token in native economics). Under Compute Exchange's published STU methodology (Kimi K2-anchored at 4.21×), GLM runs roughly 25% cooler at output than the published index. Providers absorb the basis at quote time, so a GLM-denominated commitment still prices against the standardized STU contract.

PROCURE GLM CAPACITY.

Submit a commitment request and Compute Exchange returns GLM inference quotes across the verified open-model provider network.

DISCLAIMER

GLM is an open-weight model family developed by Z.ai (Zhipu AI), released under MIT license. Compute Exchange facilitates quotes from verified third-party inference providers serving the GLM family and does not operate inference infrastructure or guarantee model availability, performance, or SLA. All commitment terms — including pricing, tap, rollover, and settlement — are negotiated directly between buyer and provider.