How to `fast` mode with OpenAI models

[rluisr]

Question

I tried it with the settings below, but compared to the codex cli, it seems that Fast mode is not enabled.
Does anyone know how to use Fast mode in OpenCode?

"openai": {
      "models": {
        "gpt-5.4": {
          "name": "GPT-5.4",
          "options": {
            "reasoningEffort": "xhigh",
            "serviceTier": "priority"
          },
          "limit": {
            "context": 400000,
            "output": 128000
          },
          "modalities": {
            "input": ["text", "image", "pdf"],
            "output": ["text"]
          },
          "variants": {
            "xhigh": {
              "serviceTier": "priority",
              "reasoningEffort": "xhigh",
              "textVerbosity": "low"
            }
          }
        }
      }
    },

[github-actions]

This issue might be a duplicate of (or related to) existing issues. Please check:

• [FEATURE]: add GPT-5.4 fast mode (/fast) as a model control #16499: Feature request for GPT-5.4 fast mode (/fast) as a TUI model control (closed)
• [FEATURE]: Support OpenAI service tier (flex/priority processing) #7511: Feature request to support OpenAI service tier (flex/priority processing) (closed)
• [FEATURE]: Support configuring OpenAI service_tier='flex' for chat/completions and document in config #12297: Feature request to configure OpenAI service_tier='flex' for chat/completions (closed)

These issues discuss enabling fast mode and serviceTier support in OpenCode for OpenAI models. Reviewing the comments and linked PRs there may answer your question or provide useful context.

[brandon-julio-t]

OpenCode already has GPT-5.4 Fast since 1.4.3 https://github.com/anomalyco/opencode/releases/tag/v1.4.3
Just make sure to use the OpenAI provider and it should show up when choosing model in the TUI.

Also what I like to do when in doubt is clone the OpenCode repo then ask opencode about the OpenCode repo.

[rluisr]

you're right......