[FEATURE]:Support full 1M context windows for models that officially provide it #39658

Activity

1. github-actions

   assigned

   jlongster

   on Jul 30, 2026
2. github-actions commented on Jul 30, 2026

   

   github-actionsBot

   on Jul 30, 2026 – with GitHub Actions

   Contributor

   More actions

   This issue might be a duplicate of existing issues. Please check:

   • [Bugs]: Context window is hardcoded to 200k for many models instead of resolving dynamically from actual provider/model metadata #35863: Context window is hardcoded to ~200k for many models instead of resolving dynamically from actual provider/model metadata (covers the same root cause — OpenCode not using the full 1M context that models like DeepSeek V4 Pro, GLM-5.2, and Qwen advertise)
   • opencode-go/glm-5.2 routed to 262K backend (DigitalOcean) despite 1M advertised context, overflow not classified #36481: opencode-go/glm-5.2 routed to 262K backend despite 1M advertised context (specifically covers GLM-5.2 and the 1M vs actual limit mismatch you described)
3. ulascanndmrc commented on Jul 30, 2026

   

   ulascanndmrc

   on Jul 30, 2026

   Author

   More actions

   I would also like to support and expand on this request.

   One of the main reasons I subscribed to OpenCode Go was because several of the supported models officially advertise extremely large context windows, such as 1M tokens for DeepSeek V4 Pro, GLM-5.2, Qwen3.7 Max, and other long-context models.

   For developers working on large, real-world software projects, the available context window is one of the most important reasons for choosing a model and paying for a subscription. Long context allows us to keep architecture documents, specifications, source code, conversations, and implementation history in a single session without constantly losing important information.

   However, OpenCode currently reports an auto-compact window of around 190K tokens. While 190K is already a large context window, it is significantly smaller than the officially advertised 1M context provided by these models. As a result, conversations are compacted much earlier than expected, making it difficult to work on large codebases over long development sessions.

   If the current limitation is caused by the backend provider, it would be very helpful if OpenCode could clearly indicate the actual available context window for the selected provider. This would allow users to understand the real limits before starting large projects.

   If the backend actually supports the full context window, I would kindly request that OpenCode expose the complete supported context instead of triggering auto-compaction much earlier.

   Even if implementing full 1M context support is not currently possible, having transparent information about the effective context limit for each provider and model would greatly improve the user experience and help users make informed decisions.

   Could you please clarify:

   • Is the ~190K auto-compact window an intentional limitation of OpenCode Go?
   • Is there any configuration that allows users to utilize the full context window supported by the underlying model?
   • Are there any plans to support the full advertised context windows (such as 1M) in the future?

   Thank you very much for your time, for considering this request, and for all the great work you continue to do on OpenCode.
4. github-actions commented on Sep 29, 2026

   

   github-actionsBot

   on Sep 29, 2026 – with GitHub Actions

   Contributor

   More actions

   To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.
5. github-actions

   closed this as not plannedon Sep 29, 2026