Runtime model metadata reporting fallback model instead of default
OpenClaw bug report — Runtime model metadata is wrong
My OpenClaw config:
• agents.defaults.model.primary = kimi-coding/k2p5
• agents.defaults.model.fallbacks = ["anthropic/claude-opus-4-6"]
When I start a new session (via /new), the agent's system prompt contains a ## Runtime line injected by the gateway:
model=anthropic/claude-opus-4-6 | default_model=kimi-coding/k2p5
The model= field shows the fallback (Opus) instead of the primary (Kimi), even though the request is actually served by Kimi. This causes the agent to incorrectly believe it's running on Opus.
Expected: model=kimi-coding/k2p5
Actual: model=anthropic/claude-opus-4-6
Is this a known issue? How does the gateway determine the model= value in the Runtime injection — does it reflect the actual model that served the request, or is it resolved from config before the API call?
My OpenClaw config:
• agents.defaults.model.primary = kimi-coding/k2p5
• agents.defaults.model.fallbacks = ["anthropic/claude-opus-4-6"]
When I start a new session (via /new), the agent's system prompt contains a ## Runtime line injected by the gateway:
model=anthropic/claude-opus-4-6 | default_model=kimi-coding/k2p5
The model= field shows the fallback (Opus) instead of the primary (Kimi), even though the request is actually served by Kimi. This causes the agent to incorrectly believe it's running on Opus.
Expected: model=kimi-coding/k2p5
Actual: model=anthropic/claude-opus-4-6
Is this a known issue? How does the gateway determine the model= value in the Runtime injection — does it reflect the actual model that served the request, or is it resolved from config before the API call?
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
131,281Members
Resources
Similar Threads
Was this page helpful?
