Vercel AI Gateway virtual models (vmc/*) only offer low, medium and high #54732

Activity

1. opencode-agent commented on Oct 12, 2026

   

   opencode-agentBot

   on Oct 12, 2026 – with OpenCode Agent

   Contributor

   More actions

   Thanks for the detailed write-up and for checking that the gateway already applies the selected effort.

   As I understand it, you'd like a configured model to point at a catalog model (your proposed base). The alias would then inherit that model's reasoning levels, limits and cost, while requests still go to the vmc/* id. Right now OpenCode can't tell what an alias resolves to, so it falls back to the generic low/medium/high list. Maintainers will consider the approach and the option name when your PR comes in.

   In the meantime, you should be able to work around it by listing the variants on the alias yourself. When variants is set, the default list isn't generated. Something like:

   "provider": {
     "vercel": {
       "models": {
         "vmc/<alias>": {
           "variants": [
             { "id": "none", "settings": { "reasoningEffort": "none" } },
             { "id": "low", "settings": { "reasoningEffort": "low" } },
             { "id": "medium", "settings": { "reasoningEffort": "medium" } },
             { "id": "high", "settings": { "reasoningEffort": "high" } },
             { "id": "xhigh", "settings": { "reasoningEffort": "xhigh" } }
           ]
         }
       }
     }
   }

   The exact settings depend on the model behind the alias, so you may need to copy them from what opencode api GET /api/model shows for the real model.

   Related:

   • feat(core): wire native Vercel AI Gateway provider #53432 and feat(ai): add native Vercel AI Gateway language models #52643: the native Vercel AI Gateway provider
   • OpenRouter route-modifier suffixes (:floor, :nitro, :exacto) can't be referenced in model IDs #48016: a similar problem with model ids that don't match the catalog (OpenRouter route suffixes)