[FEATURE]: New chat.model hook #50729

Activity

1. github-actions commented on Sep 22, 2026

   

   github-actionsBot

   on Sep 22, 2026 – with GitHub Actions

   Contributor

   More actions

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

   • [FEATURE]: add chat.model plugin hook for pre-call model routing #18793: [FEATURE]: add chat.model plugin hook for pre-call model routing (closed, same request)
   • [FEATURE]:Support runtime model switching via plugin API #24006: [FEATURE]: Support runtime model switching via plugin API (closed, same concept)
   • Plugins: chat.params hook cannot override the model (only sampling params) — model-switch plugins are forced to abort+resend, which breaks one-shot opencode run callers #49712: Plugins: chat.params hook cannot override the model — model-switch plugins are forced to abort+resend (open, directly related)
   • [FEATURE]: Add V2 model-context plugin hook #35408: [FEATURE]: Add V2 model-context plugin hook (open, related)
   • [FEATURE]: Plugin hook to intercept and override LLM call parameters (llm.request.before) #45764: [FEATURE]: Plugin hook to intercept and override LLM call parameters (open, related)
2. github-actions

   assigned

   rekram1-node

   on Sep 22, 2026
3. opencodejr commented on Oct 28, 2026

   

   opencodejrBot

   on Oct 28, 2026

   More actions

   Thanks for the request, @dmouse! You're asking for a v2 plugin hook that can pick or replace the model before each step of a turn, so a plugin can route simple steps to cheaper models and complex work to stronger ones.

   The current v2 plugin API can't do this. Session hooks such as context, model.request and http.request get the model as read-only, and it's resolved before they run. The only option a plugin has is session.switchModel, which changes the session's selected model for good rather than for one step. So this needs a change to the plugin API itself, and maintainers will consider it.

   Related issues and PRs:

   • [FEATURE] Multi-model orchestration: intent-based routing, vision delegation, turn-level advisor #34370: broader request for intent-based multi-model routing, including a routing hook for plugins
   • [FEATURE]: Add V2 model-context plugin hook #35408: the V2 model-context plugin hook (it can change system, messages and tools, but the model stays read-only)
   • feat(core): add session model.select hook for per-step model routing #50965: your closed PR adding a model.select hook for per-step routing
   • [FEATURE]: add chat.model plugin hook for pre-call model routing #18793, [FEATURE]:Support runtime model switching via plugin API #24006: the earlier chat.model requests, which were against v1