[FEATURE]: Support OpenAI service tier (flex/priority processing)

[maheshrijal]

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

OpenAI’s API supports selecting a processing/service tier. Although this can be expensive, but getting things done quickly for urgent tasks would be a great feature. Please add support for this in OpenCode.

Upstream reference: openai/codex#2916

[emamulandalib]

So far I found it was introduced 2 months ago already as part of this PR #4914 (get codex working in copilot)

opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts

Line 1706 in b267328

serviceTier: z.enum(["auto", "flex", "priority"]).nullish(),

Tested with below configuration already and it's working but yes redundant value to handle. I guess here we need a good config design.

"agent": {
    "plan": { "options": { "serviceTier": "flex" } },
    "title": { "options": { "serviceTier": "flex" } },
    "build": { "options": { "serviceTier": "flex" } }
  },

The config will be parsed here...

opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts

Lines 196 to 200 in b267328

	const openaiOptions = await parseProviderOptions({
	provider: "openai",
	providerOptions,
	schema: openaiResponsesProviderOptionsSchema,
	})

...and it will be sent to API.

opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts

Line 289 in b267328

service_tier: openaiOptions?.serviceTier,

Then, the repsonse will be captured for stream or non-stream

opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts

Lines 742 to 744 in b267328

	if (typeof response.service_tier === "string") {
	providerMetadata.openai.serviceTier = response.service_tier
	}

opencode/packages/opencode/src/provider/sdk/openai-compatible/src/responses/openai-responses-language-model.ts

Lines 1237 to 1239 in b267328

	if (typeof value.response.service_tier === "string") {
	serviceTier = value.response.service_tier
	}

Added some logs to verify

timestamp            duration  provider  model       agent  requested  response
2026-01-11T22:04:35  +1ms      openai    gpt-5.2     build  flex       flex
2026-01-11T22:04:35  +549ms    openai    gpt-5-nano  title  flex       flex
2026-01-11T22:04:38  +1006ms   openai    gpt-5.2     build  flex       flex
2026-01-11T22:04:42  +402ms    openai    gpt-5.2     build  flex       flex
2026-01-11T22:04:54  +1324ms   openai    gpt-5.2     build  flex       flex
2026-01-11T22:04:59  +122ms    openai    gpt-5.2     build  flex       flex

This works as well at model level configuration. I guess we need to update the documentation and the config schema https://opencode.ai/config.json.