#2026.4.11: Agent model defaults to openai/gpt-5.4 despite Anthropic configuration

1 messages · Page 1 of 1 (latest)

sinful dawn
#

Hi team,

After updating to OpenClaw 2026.4.11, the agent is defaulting to openai/gpt-5.4 even though my config is set to Anthropic Claude Sonnet 4.6.

Log shows:

  • anthropic/claude-sonnet-4-6 model configured, enabled automatically.
  • but agent model: openai/gpt-5.4

This happens every time I start it. It didn't happen in previous versions.

Is this expected behavior in 2026.4.11 or a bug?

Thanks!

hallow pivotBOT
brave relicBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

hallow pivotBOT
sharp walrusBOT
#

I don’t think that’s expected, if your actual default is set to agents.defaults.model.primary: "anthropic/claude-sonnet-4-6".

What is expected in 2026.4.11 is this:

  • openai/gpt-5.4 is the built-in fallback baseline
  • so if the configured default is missing, invalid, or only present in the allowlist/catalog, startup can still resolve to openai/gpt-5.4

Also, the log line like:

  • anthropic/claude-sonnet-4-6 model configured, enabled automatically

usually does not mean “this is now the active default model”. It often just means the model/provider got auto-enabled.

Can you paste these?

openclaw --version
# 
openclaw config get agents.defaults.model
openclaw config get agents.defaults.models
openclaw models status

If agents.defaults.model is empty or still OpenAI, the quick fix is:

openclaw models set anthropic/claude-sonnet-4-6

Docs:

If those commands already show Anthropic as the primary model, but startup still logs agent model: openai/gpt-5.4, then yeah, I’d treat that as a regression/bug in 2026.4.11 and ask you to file it here with the outputs plus the startup log: