OpenAI provider headers timeout after 10000ms on 1.15.11; increasing headerTimeout fixes it

[rayborg]

Description

After upgrading OpenCode via Homebrew from 1.14.28 to 1.15.11, requests using the OpenAI provider began failing with:

Provider response headers timed out after 10000ms

Increasing the OpenAI provider headerTimeout resolved the issue locally.

This looks like a regression or a default timeout that may be too aggressive for some OpenAI/ChatGPT-provider requests or network environments.

Plugins

No custom plugins involved in the repro.

OpenCode version

1.15.11

Installed via Homebrew.

Steps to reproduce

1. Upgrade OpenCode from 1.14.28 to 1.15.11 via Homebrew.
2. Use the OpenAI provider with gpt-5.5-fast / GPT-5.5 Fast.
3. Run a normal request.
4. Observe the request fail with:

Provider response headers timed out after 10000ms

5. Add the following provider options to the global config:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "openai": {
      "options": {
        "headerTimeout": 60000,
        "timeout": 600000,
        "chunkTimeout": 60000
      }
    }
  }
}

6. Verify with opencode debug config that the options are loaded.
7. Run the same request again.
8. Observe that it succeeds after increasing headerTimeout.

Screenshot and/or share link

N/A. Relevant terminal error is included above.

Operating System

macOS, arm64

Terminal

zsh

Question

Is the 10000ms default intended for OpenAI/ChatGPT-provider requests in 1.15.11? It may be too low for some responses or network/proxy environments.

[github-actions]

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

• Desktop hits 5-minute Headers Timeout Error with slow local providers #26602: Same Headers Timeout Error with default timeout being too low for some provider requests; same workaround of increasing headerTimeout/timeout in config.
• 1.2.25 timeouts are too aggressive for larger local models #17307: Same pattern of timeouts becoming too aggressive after an upgrade, with the same fix of increasing timeout values in opencode.json provider options.

[rayborg]

Clarifying the duplicate suggestions:

• Desktop hits 5-minute Headers Timeout Error with slow local providers #26602 is related, but it describes Desktop/local OpenAI-compatible providers timing out after about 5 minutes (Headers Timeout Error) while waiting on slow local providers such as Ollama.
• 1.2.25 timeouts are too aggressive for larger local models #17307 is also related, but it focuses on local/larger model timeouts around 1.2.25 and longer-running prompt/stream timeouts.

This issue is specifically a shorter 10000ms header timeout observed after upgrading to 1.15.11 with the OpenAI provider / ChatGPT-provider path using gpt-5.5-fast.

The practical difference is the failing threshold and provider path:

• This issue: OpenAI provider, Provider response headers timed out after 10000ms, fixed by setting provider.openai.options.headerTimeout to 60000.
• Desktop hits 5-minute Headers Timeout Error with slow local providers #26602: local OpenAI-compatible provider / Desktop path, timeout around 5 minutes.
• 1.2.25 timeouts are too aggressive for larger local models #17307: local models / larger context timeout behavior after an older upgrade.

So this may share the broader timeout-config surface, but it seems worth tracking separately unless maintainers consider the 1.15.11 OpenAI 10000ms default intentionally covered by those issues.

[rekram1-node]

This is intentional, many people are getting no response from openai for extended periods of time almost every other request it seems for certain folks, so we have this header timeout to account for that, as seen in the release notes for the release.

We may bump it but it's entirely configurable and u can even disable it by setting headerTimeout: false

[C485]

This is intentional, many people are getting no response from openai for extended periods of time almost every other request it seems for certain folks, so we have this header timeout to account for that, as seen in the release notes for the release.

We may bump it but it's entirely configurable and u can even disable it by setting headerTimeout: false

i guess we need something that will disable all timeouts in config file, for now
"timeout": 3000000, "chunkTimeout": 3000000, "headerTimeout": false
does nothing at all, for local providers we're still getting
TypeError: terminated at Fetch.onAborted (node:internal/deps/undici/undici:13602:53)
I would be great to have it as soon as possible

[rekram1-node]

@C485 this seems like something else entirely none of those timeouts are enabled by default besides header timeout but its only enabled for openai provider and thats it (by default)

[Dayclone]

I get "Provider response headers timed out after 10000ms" now as well when it never used to happen

[rekram1-node]

@Dayclone u can bump it

[Dayclone]

@rekram1-node I did
"provider": { "openai": { "options": { "timeout": 20000 } },

But probably did that wrong?

[rekram1-node]

Yes u did, read above msgs, it is headerTimeout

[Dayclone]

Yep that was it thank you! Seems to be working now.

[Abdo-Eid]

From what I noticed, the problem appears after GPT models read/analyze images and disappears when I compact the conversation or talk in a new chat
i disabled the headerTimeout as @rekram1-node suggested, it took some time but it eventually responded without compaction
@rayborg