# Atlas with GLM-5: Frontier AI Coding in 2026 > GLM-5 from Z.ai, released in February 2026, provides a 204,800 token context window and costs $1.00 per Mtok for input. GLM-5 is Z.ai's frontier model for Atlas in 2026, offering a 204,800 token context window and advanced reasoning at an input cost of $1.00 per Mtok. It is ideal for developers needing to manage large codebases and perform significant rewrites within Atlas, balancing powerful capabilities with a competitive price point compared to other frontier options. ## Key takeaways - GLM-5 offers a 204,800 token context window, suitable for large service and test analysis. - Input pricing is $1.00 per Mtok, making it a cost-effective frontier model in 2026. - Output pricing is $3.20 per Mtok, which can be optimized by using a Flash tier for execution. - It is the base of the GLM-5 line, sharing tooling with GLM-5.1 and GLM-5.2. - GLM-5.2 provides 5x the context (1,000,000 tokens) for an additional $0.40 per Mtok input. - The API is China-hosted, requiring a gateway for US/EU data residency. ## What is GLM-5 best for in Atlas? GLM-5, Z.ai's frontier model released in February 2026, excels within Atlas for tasks requiring deep code understanding and extensive modifications, thanks to its 204,800 token context window. This model is particularly suited for developers needing to manage large services and their associated tests. Atlas, driven by GLM-5, can effectively handle complex coding challenges. Its 204,800 token context window, coupled with a 131,072 token maximum output, allows Atlas to load an entire large service along with its tests, enabling comprehensive analysis and full rewrites. This capability is crucial when Atlas uses its Axis, the hybrid semantic and keyword code search, to retrieve relevant code, or when it drafts a plan in its read-only plan agent. The model's frontier-tier reasoning capabilities ensure that Atlas can generate sophisticated plans and execute detailed code changes, presenting a unified diff for approval before writing. For developers in 2026, GLM-5 provides a powerful engine for Atlas to perform significant refactoring, feature development, and bug fixes across substantial codebases. ## What are the cost and context tradeoffs of GLM-5? GLM-5 offers a compelling balance of cost and context, priced at $1.00 per Mtok input and $3.20 per Mtok output, with a 204,800 token context window. This positions it as a cost-effective frontier reasoning model in 2026, especially when compared to more expensive alternatives like GLM-5.1 and GLM-5.2. While GLM-5 provides a generous 204,800 token context window, a key tradeoff exists with its sibling, GLM-5.2, which boasts a 1,000,000 token context. This 5x larger context in GLM-5.2 comes at an additional $0.40 per Mtok input, making it $1.40 per Mtok input. For developers working on extremely large repositories where the entire codebase might exceed GLM-5's capacity, the jump to GLM-5.2 could be justified despite the increased cost. However, for many large service-level tasks, GLM-5's 204,800 tokens are sufficient. The $1.00 per Mtok input price for GLM-5 is notably lower than GLM-5.1's $1.40 and significantly less than Western frontier pricing, making it an attractive option for budget-conscious teams in 2026. Atlas's ability to switch the active model on the fly allows developers to leverage GLM-5 for most tasks and only switch to a higher-context model when absolutely necessary. ## When should I choose a different model over GLM-5? While GLM-5 is a strong frontier contender in 2026, developers should consider alternative models for specific scenarios, particularly when dealing with extremely large repositories or strict data residency requirements. Its 204,800 token context, while substantial, is not the largest available within the GLM-5 family. Developers should consider a different model if their projects consistently require a context window significantly larger than GLM-5's 204,800 tokens. For instance, GLM-5.2 offers a 1,000,000 token context, which is roughly 5 times larger, making it more suitable for whole-repository analysis or tasks that span an exceptionally broad scope. This comes at a higher input cost of $1.40 per Mtok, an increase of $0.40 over GLM-5. Another critical factor is data residency. GLM-5's API is China-hosted. If your organization has US or EU data-residency requirements, you would need to implement a gateway solution, or opt for a model from a different provider that meets those geographical compliance standards. Atlas's flexibility to switch models on the fly, with favorites and recents, means you can easily configure a "small_model" like "zai/glm-4.7-flash" for background tasks to optimize costs, while reserving GLM-5 for complex reasoning, or switch to GLM-5.2 for massive context needs. ## Setup 1. 1: Set your Z.ai API Key: Export your Z.ai API key by running `export ZHIPU_API_KEY="YOUR_API_KEY_HERE"` in your terminal, or use `atlas login` and select Z.ai as your provider. 2. 2: Verify GLM-5 resolution: Confirm that `glm-5` resolves correctly from the models.dev registry by executing the command `atlas models zai`. 3. 3: Pin GLM-5 in atlas.json: Configure Atlas to use GLM-5 by adding `"model": "zai/glm-5"` to your `atlas.json` configuration file. 4. 4: Optimize with a small model: To manage costs for background operations, also pin a smaller, more economical model like `"small_model": "zai/glm-4.7-flash"` in your `atlas.json`. 5. 5: Delegate execution to Flash tier: Allow Atlas's plan agent to utilize GLM-5 for its reasoning and planning, then hand off execution tasks to a Flash tier model to keep the $3.20/Mtok output rate focused on the strategic thinking rather than mechanical operations. ## FAQ ### What is the context window size for GLM-5 in Atlas? GLM-5 provides a substantial 204,800 token context window, allowing Atlas to process and generate code for large services and their associated tests effectively. ### How much does it cost to use GLM-5 with Atlas? Using GLM-5 with Atlas costs $1.00 per Mtok for input tokens and $3.20 per Mtok for output tokens, making it a competitive frontier model in 2026. ### Can Atlas use GLM-5 for full code rewrites? Yes, with its 204,800 token context and 131,072 token max output, GLM-5 enables Atlas to hold a large service plus its tests and emit a full rewrite, presenting a unified diff for approval. ### What are the data residency implications of using GLM-5? GLM-5's API is China-hosted. If your organization has US or EU data-residency requirements, you will need to route traffic through a compliant gateway. ### How does GLM-5 compare to GLM-5.2 for context window? GLM-5 has a 204,800 token context, while GLM-5.2 offers a significantly larger 1,000,000 token context, roughly 5 times more, for an additional $0.40 per Mtok input. ### How can I optimize costs when using GLM-5 with Atlas? To optimize costs, configure Atlas to use GLM-5 for planning and reasoning, and then delegate execution tasks to a more economical Flash tier model like GLM-4.7-Flash, reducing the $3.20/Mtok output rate on mechanical operations. ### Is GLM-5 compatible with other GLM-5 family models in Atlas? Yes, GLM-5 is the base of the GLM-5 line and shares tooling and prompt conventions with models like GLM-5-Turbo and GLM-5.1, allowing for drop-in ID swaps in your `atlas.json` configuration. ### What is the release date of GLM-5? GLM-5 was released in February 2026 by Z.ai, marking a generational jump past their 4.x model line. --- Canonical HTML: https://seashell.sh/resources/models/glm-5 Source of truth: aeo_pages row `/resources/models/glm-5` (segment: Models) (this file is generated from it, never hand-edited). Licence: SeaShell is proprietary with a free core. It is not open source and there is no public source repository.