Muse Glimmer
Meta's open-weight 30B agentic model that runs local, multimodal AI agents on a single GPU
Open-weight frontier LLM family from Z.ai (Zhipu AI), tuned for coding and agents.
One of the strongest open-weight coding and agentic model families, with genuinely permissive licensing and low API costs, though the very newest releases (like GLM-5.3) may reach coding plans before weights or API pricing are published.
GLM is the open-weight LLM family from Z.ai (Zhipu AI), with the GLM-5.x series positioned as a frontier open model for agentic coding and reasoning. It offers a free chat product at chat.z.ai, openly licensed weights on Hugging Face for self-hosting, and a low-cost API plus a coding subscription plan. It is widely used as an open-source alternative to closed frontier models.
GLM (General Language Model) is the model family built by Z.ai, the international brand of Beijing-based Zhipu AI, a Tsinghua University spin-off and the first publicly listed large language model company (HKEX: 2513). The 2026 flagship GLM-5.x series is a large Mixture-of-Experts system aimed at agentic coding, long-context reasoning, and tool use, with weights for releases like GLM-5.2 published openly under an MIT license so developers can self-host, fine-tune, and use them commercially. Users can try GLM for free through the chat.z.ai web app, while developers integrate the models through Z.ai's OpenAI-compatible API or a dedicated GLM Coding Plan that plugs into tools such as Claude Code, Cline, and OpenCode. This mix of open weights, competitive token pricing, and coding-tool integrations makes GLM a popular open-source alternative to closed frontier models for teams that want control over deployment and cost.
GLM is the open-weight large language model family from Z.ai, the international brand of Beijing-based Zhipu AI. The 2026 GLM-5.x flagship is a large Mixture-of-Experts model aimed at agentic coding and long-context reasoning. Users get free chat at chat.z.ai, developers get a low-cost OpenAI-compatible API and a coding plan, and major releases publish MIT-licensed weights on Hugging Face. It is a leading open-source alternative to closed frontier models.
GLM is developed by Zhipu AI, which markets its models and products internationally under the Z.ai brand at z.ai. Zhipu AI was founded in 2019 as a spin-off from the Knowledge Engineering Group at Tsinghua University and is headquartered in Beijing. It has raised substantial capital from investors including Alibaba, Tencent, Meituan, Ant Group, Xiaomi, HongShan, and Prosperity7 Ventures, and it became the first publicly listed large language model company through a Hong Kong IPO (HKEX: 2513).
The company builds the GLM (General Language Model) family alongside multimodal models and agent systems, and it has pursued an open-weight strategy that publishes model weights under permissive licenses. This positions Z.ai as both a commercial API provider and a major contributor to the open-source AI ecosystem.
The GLM-5.x series is a large Mixture-of-Experts architecture (the GLM-5.2 flagship is around 744B total parameters with roughly 40B active per token) built for agentic coding, tool use, and long-context reasoning, with very large context windows on the flagship releases. Weights for major releases such as GLM-5.2 are openly available on Hugging Face (zai-org) and ModelScope under the MIT license, enabling self-hosting, fine-tuning, and commercial deployment.
Beyond raw models, Z.ai delivers GLM through several channels: a free chat interface at chat.z.ai, an OpenAI-compatible API with competitive per-token pricing and free Flash tiers, and a GLM Coding Plan that integrates the models into agentic coding tools like Claude Code, Cline, and OpenCode. This combination lets teams choose between hosted convenience and full self-managed control.
GLM primarily targets developers, engineering teams, and AI researchers who want a high-performing open-weight model for coding assistants, autonomous agents, and long-context applications, especially those seeking self-hosting, fine-tuning, or lower API costs than closed frontier providers. It also serves cost-sensitive startups and enterprises evaluating open-source alternatives, as well as individual users who simply want free access to a capable chatbot at chat.z.ai.
Software developers, AI engineers, and researchers who use GLM to build coding assistants and agents, run long-context reasoning, or fine-tune open weights; plus casual users chatting for free at chat.z.ai.
Engineering leads, CTOs, and technical founders who decide on model providers, weighing open-weight control, API cost, licensing, and integration with existing coding tools.
Open-source AI communities, ML practitioners on Hugging Face, benchmark evaluators, and developer-tool ecosystems (Claude Code, Cline, OpenCode) that shape adoption.
A developer-led team or AI-focused company that wants a top-tier open-weight coding and agentic model it can self-host, fine-tune, or access cheaply via API, and that values permissive MIT licensing and integration with agentic coding tools over a fully closed single-vendor SaaS.
Zhipu AI (Z.ai) has raised substantial funding, including a roughly RMB 2.5B round in 2023 backed by Alibaba, Tencent, Meituan, Ant Group, Xiaomi, and HongShan, and a ~$400M round in 2024 led by Prosperity7 Ventures, plus later state-linked investment. It became the first publicly listed LLM company via a Hong Kong IPO (HKEX: 2513) in early 2026.
Yes, in part. You can chat with GLM for free at chat.z.ai, and open weights for major releases like GLM-5.2 are free to download and self-host under the MIT license. Higher-volume use through the Z.ai API is usage-based, and there is a separate paid GLM Coding Plan.
GLM competes directly with other leading open-weight Chinese model families like DeepSeek, Alibaba's Qwen, and Moonshot's Kimi. The GLM-5.x series is positioned by Z.ai as one of the strongest open-weight models for coding and agentic tasks; the best choice depends on your benchmarks, licensing needs, and tooling, and you should validate performance on your own workloads rather than rely on vendor claims.
The weights for major GLM releases, such as GLM-5.2, are published openly on Hugging Face and ModelScope under a permissive MIT license, allowing self-hosting, fine-tuning, and commercial use. Note that some brand-new releases (for example GLM-5.3) may launch in products first and open their weights later after review.
Z.ai's API is usage-based. Verified rates include roughly $1.40 per million input tokens and $4.40 per million output tokens for GLM-5.2, and about $0.60 / $2.20 for the cheaper GLM-4.7, with some Flash models offered free. There is also a GLM Coding Plan at Lite $18, Pro $72, and Max $160 per month.
GLM is made by Zhipu AI, which operates internationally under the Z.ai brand. Zhipu AI is a Beijing-based frontier AI lab founded in 2019 as a spin-off from Tsinghua University, and it became the first publicly listed large language model company via a Hong Kong IPO (HKEX: 2513).
Side-by-side pages for pricing, features, and best-fit use cases.
This week in AI: Meta's open-weight Muse Glimmer, an escalating model price war, Perplexity Comet goes free, OpenAI retires Atlas, and a wave of inference funding.
Local AI went mainstream in 2026. A developer's guide to the best open-weight models to run on your own GPU or Mac: Muse Glimmer, Gemma, Qwen, DeepSeek, and GLM.
Open-weight Chinese models now rival frontier labs on code for a fraction of the price. We compare DeepSeek, Qwen, Kimi and GLM on licensing, access, cost and real coding strength.
Meta's open-weight 30B agentic model that runs local, multimodal AI agents on a single GPU
Alibaba's free AI assistant, backed by the open-weight Qwen model family
Google's family of open-weight AI models you can download, run locally, and self-host
Free AI chat assistant from Moonshot AI, built for long-context reasoning and open Kimi K2 models