Issue Station
Sign In
Tools hereToolbox

openclaw - ✅(Solved) Fix gpt-5-mini returns 400: reasoning_effort 'none' not supported (since v2026.4.5) [2 pull requests, 2 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
openclaw/openclaw#62967•Fetched 2026-04-09 08:00:01
View on GitHub
Comments
2
Participants
2
Timeline
10
Reactions
3
Author
weiyong1024weiyong1024
Participants
dhqdbbdhqdbbweiyong1024weiyong1024
Timeline (top)
cross-referenced ×4commented ×2referenced ×2subscribed ×2

Error Message

  1. → 400 error, no response
  • #54844 — gpt-5-mini 400 error via github-copilot provider (possibly same root cause)

Root Cause

  • #33272 — same mechanism (thinkingDefault → reasoning_effort) but for custom/Ollama providers
  • #54844 — gpt-5-mini 400 error via github-copilot provider (possibly same root cause)

Fix Action

Workaround

Setting agents.defaults.thinkingDefault to "minimal" and restarting the gateway resolves the issue:

openclaw config set agents.defaults.thinkingDefault minimal

PR fix notes

PR #63048: fix: stop sending unsupported reasoning_effort 'none' to OpenAI

Description (problem / solution / changelog)

Summary

Fixes #62967.

Problem

When a reasoning-capable model like gpt-5-mini was used without an explicit thinkingDefault, the OpenAI Responses API transport hardcoded reasoning: { effort: "none" } in the request payload. gpt-5-mini rejects "none" with a 400 error — it only accepts "minimal", "low", "medium", and "high".

Solution

Remove the fallback that forces effort: "none". When no reasoning effort is explicitly configured, the reasoning block is now omitted entirely, letting the API use its own default.

Changes

  • src/agents/openai-transport-stream.ts: Remove the else if branch that sent { effort: "none" }, and clean up the corresponding type union.

Testing

  • Existing tests pass (40/40)
  • Matches the workaround described in the issue (omit reasoning when not configured)

<sub>🔧 Generated by issue-to-pr</sub>

Changed files

  • src/agents/openai-transport-stream.ts (modified, +4/-8)

PR #85: chore: bump OpenClaw to 2026.4.9

Description (problem / solution / changelog)

Summary

  • Update RecommendedOpenClawVersion from 2026.4.1 to 2026.4.9
  • Fixes reasoning_effort:'none' bug that broke gpt-5-mini on versions 4.5–4.7 (openclaw/openclaw#62967)

Test plan

  • Image build with 2026.4.9
  • Instance create + configure (gpt-5-mini)
  • Gateway health check pass
  • Chat compatibility — no session errors (reasoning_effort bug fixed)
  • Control Panel chat — user verified bot responds normally

🤖 Generated with Claude Code

Changed files

  • internal/version/version.go (modified, +1/-1)

Code Example

400 Unsupported value: 'none' is not supported with the 'gpt-5-mini' model.
Supported values are: 'minimal', 'low', 'medium', and 'high'.

---

openclaw config set agents.defaults.thinkingDefault minimal
RAW_BUFFERClick to expand / collapse

Bug Description

Since OpenClaw v2026.4.5 (also confirmed in v2026.4.7), sending a message to gpt-5-mini via OpenAI's native API fails with:

400 Unsupported value: 'none' is not supported with the 'gpt-5-mini' model.
Supported values are: 'minimal', 'low', 'medium', and 'high'.

OpenClaw is sending reasoning_effort: "none" in the API request, but OpenAI's gpt-5-mini does not accept "none" — only "minimal", "low", "medium", and "high".

Steps to Reproduce

  1. Fresh install of OpenClaw v2026.4.5 or v2026.4.7
  2. openclaw onboard with an OpenAI API key
  3. openclaw models set openai/gpt-5-mini
  4. Start gateway, open Control Panel, send any message
  5. → 400 error, no response

Expected Behavior

gpt-5-mini should respond normally. If reasoning_effort is not explicitly configured by the user, OpenClaw should either omit the parameter or use a valid default (e.g. "low").

Workaround

Setting agents.defaults.thinkingDefault to "minimal" and restarting the gateway resolves the issue:

openclaw config set agents.defaults.thinkingDefault minimal

Environment

  • OpenClaw: v2026.4.5 and v2026.4.7 (works fine on v2026.4.1)
  • Provider: OpenAI (native API, not custom/proxy)
  • Model: gpt-5-mini
  • Platform: Docker container (linux/arm64), also reproduced on linux/amd64

Related Issues

  • #33272 — same mechanism (thinkingDefault → reasoning_effort) but for custom/Ollama providers
  • #54844 — gpt-5-mini 400 error via github-copilot provider (possibly same root cause)

extent analysis

TL;DR

Setting agents.defaults.thinkingDefault to a supported value, such as "minimal", resolves the issue with sending messages to gpt-5-mini via OpenAI's native API.

Guidance

  • Verify that the reasoning_effort parameter is being sent with the value "none" in the API request, which is not supported by the gpt-5-mini model.
  • Update the thinkingDefault configuration to a valid value (e.g., "minimal", "low", "medium", or "high") using the command openclaw config set agents.defaults.thinkingDefault <value>.
  • Restart the gateway after updating the configuration to apply the changes.
  • Test sending a message to gpt-5-mini again to confirm that the issue is resolved.

Example

openclaw config set agents.defaults.thinkingDefault minimal

Notes

This fix assumes that the issue is solely due to the unsupported reasoning_effort value. If the problem persists, further investigation into the OpenClaw configuration and OpenAI API interaction may be necessary.

Recommendation

Apply the workaround by setting agents.defaults.thinkingDefault to a supported value, as this resolves the issue without requiring an upgrade to a potentially unavailable fixed version.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#api#dependency conflict#environment setup#docker error#permission error

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING