Fix "does not allow /v1/messages dispatch" and /v1/messages 400s

Why Claude Code or the Anthropic SDK gets "This group does not allow /v1/messages dispatch" or a 400 on /v1/messages, and which models and settings fix it on BigModel.io.

On this page
  1. What it means
  2. How to fix it
  3. Troubleshooting
  4. Related

Claude Code and the Anthropic SDKs call POST /v1/messages. Two errors mean that this endpoint is not available for the model or key you used:

json
{ "type": "error", "error": { "type": "permission_error", "message": "This group does not allow /v1/messages dispatch" } }
json
{ "type": "error", "error": { "type": "invalid_request_error", "message": "`X` does not support the Anthropic Messages API (/v1/messages, used by Claude Code) yet. Models that do: ... Or call /v1/chat/completions or /v1/responses." } }

#What it means

/v1/messages serves only models whose endpoints include it. Today those are glm-5.3, kimi-k3 and deepseek-v4-flash. Every text model works on /v1/chat/completions, and the model catalog lists the endpoints of each model.

  • The 400 names the models that do support the Messages API. Pick one of them.
  • The 403 "does not allow /v1/messages dispatch" only appears when a request is not matched to a catalog model. On BigModel.io you normally get the 400 above, or a 404 for an unknown model ID. Using a supported model ID fixes both.

If you saw this 403 on another service, its operator has to enable Messages for your key there.

#How to fix it

  1. Set the model to one that supports /v1/messages. In Claude Code, set all model slots, because background tasks use the haiku slot:
bash
export ANTHROPIC_BASE_URL="https://bigmodel.io"
export ANTHROPIC_AUTH_TOKEN="$BIGMODEL_API_KEY"
export ANTHROPIC_MODEL="glm-5.3"
export ANTHROPIC_DEFAULT_OPUS_MODEL="glm-5.3"
export ANTHROPIC_DEFAULT_SONNET_MODEL="glm-5.3"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="deepseek-v4-flash"
  1. Check the base URL: https://bigmodel.io without /v1. Claude Code adds /v1/messages itself.
  2. To use a model without Messages support, call it from an OpenAI-compatible client on https://bigmodel.io/v1/chat/completions instead.

#Troubleshooting

The error names a claude-* model. Claude Code fell back to its default model because a slot is unset. Set every ANTHROPIC_DEFAULT_*_MODEL value. See Model not found.

It worked yesterday with the same model. Check the changelog and the model's page for endpoint changes, then contact Support with the X-Request-ID.

Updated · View as Markdown