# GLM-5
GLM-5 is a 744-billion-parameter open-weight Mixture-of-Experts (MoE) large language model released by the Chinese AI lab [Zhipu AI](https://www.edgepedia.org/zhipu-ai) under its international brand [Z.ai](https://www.edgepedia.org/z-ai) on February 11, 2026, with 40 billion parameters active per token.[1](https://awesomeagents.ai/models/glm-5/) • [2](https://arxiv.org/html/2602.15763v2) Z.ai positions it for complex systems engineering and long-horizon agentic tasks, meaning multi-step work in which a model plans, writes and executes code, and operates tools over extended sessions.[3](https://z.ai/blog/glm-5) This article covers the GLM-5 release itself; the GLM family, Zhipu AI, and the [GLM-5.3-Flash](https://www.edgepedia.org/glm-5-3-flash) variant have their own articles.
| Fact | Detail |
|---|---|
| Release date | February 11, 2026[1](https://awesomeagents.ai/models/glm-5/) |
| Size | 744B total parameters, 40B active (MoE, 256 experts with 8 active per token, 80 layers)[2](https://arxiv.org/html/2602.15763v2) |
| Training data | 28.5 trillion tokens (up from GLM-4.5's 23T)[3](https://z.ai/blog/glm-5) |
| Context window | 200K tokens, extended to 202,752 during SFT; 1M tokens in GLM-5.2[2](https://arxiv.org/html/2602.15763v2) • [4](https://github.com/zai-org/glm-5) |
| License | MIT, weights on Hugging Face, and ModelScope[3](https://z.ai/blog/glm-5) |
| API pricing | $1.00 per million input tokens, $3.20 per million output tokens[1](https://awesomeagents.ai/models/glm-5/) |
| Independent score | 50 on the Artificial Analysis Intelligence Index v4.0, the first open-weights model to reach 50[2](https://arxiv.org/html/2602.15763v2) |
## What GLM-5 is
GLM-5 doubles the scale of its predecessor GLM-4.5, which had 355 billion total and 32 billion active parameters. The technical report describes a model with 256 experts, of which 8 are active per token, arranged in 80 layers; the report states that the layer count was reduced to minimize expert-parallelism communication overhead.[2](https://arxiv.org/html/2602.15763v2) Z.ai's launch post frames the release as a move "from vibe coding to agentic engineering," targeting long-horizon agentic workloads rather than single-turn code generation.[3](https://z.ai/blog/glm-5)
The 2026 point releases ship as a family in both BF16 and FP8 weight formats: GLM-5, GLM-5.1, GLM-5.2, and GLM-5.3 at 744B-A40B, plus a smaller GLM-5.3-Flash at 320B total and 18B active parameters.[4](https://github.com/zai-org/glm-5) GLM-5.2, described in the official repository as the flagship for long-horizon tasks, is the first GLM model to deliver a 1M-token context window.[4](https://github.com/zai-org/glm-5)
## Architecture and training as published
The published specifications come from Zhipu's technical report and launch materials, and should be read as vendor disclosures rather than independently audited ones.[2](https://arxiv.org/html/2602.15763v2)
**Scale and attention.** The base model was trained on 28.5 trillion tokens, starting from a 27-trillion-token initial corpus that prioritized code and reasoning; mid-training then extended the context window from 4K to 200K tokens.[2](https://arxiv.org/html/2602.15763v2) GLM-5 adopts DeepSeek Sparse Attention (DSA), a sparse-attention mechanism introduced by DeepSeek, which Z.ai says significantly reduces deployment cost while preserving long-context capacity.[2](https://arxiv.org/html/2602.15763v2) • [3](https://z.ai/blog/glm-5) [Supervised fine-tuning](https://www.edgepedia.org/supervised-fine-tuning) extended the maximum context length to 202,752 tokens.[2](https://arxiv.org/html/2602.15763v2)
**Reasoning and post-training.** GLM-5 is a single hybrid model rather than separate reasoning and non-reasoning variants: it supports interleaved thinking before every response and every tool call, plus turn-level thinking modes that let applications control how much deliberation occurs.[2](https://arxiv.org/html/2602.15763v2) Post-training runs a sequential reinforcement-learning pipeline, moving from Reasoning RL to Agentic RL to General RL, with On-Policy Cross-Stage Distillation, built on a new asynchronous RL infrastructure that decouples generation from training.[2](https://arxiv.org/html/2602.15763v2) Z.ai's developer docs name this framework Slime, built to support larger model scales and more complex reinforcement-learning tasks.[5](https://docs.z.ai/guides/llm/glm-5)
**Point releases.** GLM-5.2 introduces IndexShare, which reuses the same indexer across every four sparse-attention layers, reducing per-token FLOPs by 2.9x at a 1M-token context length.[4](https://github.com/zai-org/glm-5) GLM-5.3-Flash is the first GLM with a hybrid sparse-and-linear attention architecture, adds Manifold-Constrained Hyper-Connections (mHC), and was pre-trained on a 30-trillion-token multimodal corpus.[4](https://github.com/zai-org/glm-5)
## Benchmark results: vendor versus independent
Vendor-reported scores. Zhipu's launch materials report that GLM-5 (Thinking) achieves 30.5 on [Humanity's Last Exam](https://www.edgepedia.org/humanitys-last-exam) (50.4 with tools), 92.7 on AIME 2026 I, 86.0 on GPQA-Diamond, and 77.8 on SWE-bench Verified.[3](https://z.ai/blog/glm-5) Across eight agentic, reasoning, and coding benchmarks, the technical report claims an average improvement of about 20% over GLM-4.7, and states that GLM-5 is "comparable to Claude Opus 4.5 and GPT-5.2 (xhigh), and better than Gemini 3 Pro."[2](https://arxiv.org/html/2602.15763v2) Z.ai's docs claim open-source state of the art in coding and agent capabilities, with real-programming usability approaching [Claude Opus 4.5](https://www.edgepedia.org/claude-opus-4-5).[5](https://docs.z.ai/guides/llm/glm-5) For the point releases, the repository reports GLM-5.2 at 81.0 on [Terminal-Bench](https://www.edgepedia.org/terminal-bench) 2.1 and 62.1 on SWE-bench Pro, up from 62.0 and 58.4 for GLM-5.1, and within a few points of Claude Opus 4.8's 85.0 on Terminal-Bench 2.1.[4](https://github.com/zai-org/glm-5)
Independent measurement. The one independent evaluation cited in available sources is the Artificial Analysis Intelligence Index v4.0, where GLM-5 scores 50, up from GLM-4.7's 42, driven by gains in agentic performance and knowledge/hallucination measures; the index reports this as the first time an open-weights model has reached 50 on that scale.[2](https://arxiv.org/html/2602.15763v2) Independent journalism has treated the parity claims differently: The Decoder's February 2026 coverage reports the 744B/40B specifications and MIT license as fact but frames the claimed parity with top Western models as Zhipu's own claim rather than a verified result.[6](https://the-decoder.com/chinese-ai-lab-zhipu-releases-glm-5-under-mit-license-claims-parity-with-top-western-models/) As of September 2026, parity with Claude Opus 4.5 and GPT-5.2 remains a vendor claim supported by only thin independent evaluation.
## Licensing, availability and pricing
GLM-5's weights are released under the [MIT License](https://www.edgepedia.org/mit-license) and are hosted on [Hugging Face](https://www.edgepedia.org/hugging-face) and [ModelScope](https://www.edgepedia.org/modelscope).[3](https://z.ai/blog/glm-5) The model is served through Z.ai's developer platform api.z.ai and through BigModel.cn, with compatibility with Claude Code and OpenClaw, meaning developers can point existing agentic coding tools at GLM-5 as a drop-in backend.[3](https://z.ai/blog/glm-5)
Listed API pricing is $1.00 per million input tokens and $3.20 per million output tokens, a blended rate of roughly $1.55 per million tokens at a 3:1 input/output ratio.[1](https://awesomeagents.ai/models/glm-5/) Third-party comparison material lists GLM-5 as undercutting [Claude Opus 4](https://www.edgepedia.org/claude-opus-4).6 by 5x and GPT-5.2 by 15x on input price.[1](https://awesomeagents.ai/models/glm-5/) No source in the available evidence provides head-to-head capability or price figures against DeepSeek, Qwen, or Kimi specifically.
## What GLM-5 signals about Chinese open-weight models
Three features of the release mark how the position of Chinese open-weight models has shifted since late 2023, when open Chinese models were generally viewed as followers of Western frontier systems.
**First past a threshold.** GLM-5 is the first open-weights model to reach 50 on the Artificial Analysis Intelligence Index v4.0, an independent aggregate measure of model capability, jumping 8 points from GLM-4.7's 42.[2](https://arxiv.org/html/2602.15763v2) A permissive MIT license on a model at this capability level means any company, including US and EU firms, can download and build on weights without negotiating with the vendor.
**Capital and hardware context.** The release followed Zhipu AI's $558 million Hong Kong IPO in January 2026, which made it China's first publicly traded AI-native company.[1](https://awesomeagents.ai/models/glm-5/) A third-party model profile reports that GLM-5 was trained on 100,000 [Huawei Ascend](https://www.edgepedia.org/huawei-ascend) 910B chips using the MindSpore framework.[1](https://awesomeagents.ai/models/glm-5/) This claim, if confirmed, would bear directly on export-control debates, since it would suggest frontier-scale training on domestic Chinese accelerators rather than US-exported GPUs; neither Zhipu nor any independent source in the available evidence confirms it.
## References
1. [GLM-5 | Awesome Agents](https://awesomeagents.ai/models/glm-5/)
2. [GLM-5: from Vibe Coding to Agentic Engineering (technical report)](https://arxiv.org/html/2602.15763v2)
3. [GLM-5: From Vibe Coding to Agentic Engineering (Z.ai launch post)](https://z.ai/blog/glm-5)
4. [zai-org/GLM-5 (official repository)](https://github.com/zai-org/glm-5)
5. [GLM-5 (Z.ai developer docs)](https://docs.z.ai/guides/llm/glm-5)
6. [Chinese AI lab Zhipu releases GLM-5 under MIT license, claims parity with top Western models (The Decoder)](https://the-decoder.com/chinese-ai-lab-zhipu-releases-glm-5-under-mit-license-claims-parity-with-top-western-models/)
---
*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › Model families and named models › Large language model families › Open-weight model families*
*Initially written Sep 17, 2026 · Edited: Oct 11, 2026; Oct 15, 2026*
*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*
License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license