issue with a LoRa finetune on the Workers AI platform
The REST API for running inference (/run) fails with a "model does not exist" error, even though the management API (/finetunes) correctly lists the finetune as existing under the same credentials.
1. Proof the Finetune Exists (Successful GET Request):
This curl command, using my API token, correctly lists the finetune, proving it exists and my credentials are valid.
curl -X GET "https://api.cloudflare.com/client/v4/accounts/ACCOUNT_ID/ai/finetunes" \
-H "Authorization: Bearer {token}
Result:
{"success":true,"result":[{"id":"214de035-8495-4c51-af6f-6687a2c29fdd","name":"my-gemma-finetune", ...}]}
2,2. Proof the Inference Call Fails (Failing POST Request):
This curl command, using the exact same credentials and the ID from the successful call above, fails.
curl -X POST "https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/run/@cf/google/gemma-3-12b-it" \
-H "Authorization: Bearer {token} \
-d '{"messages":[{"role":"user","content":"..."}],"lora":"214de035-8495-4c51-af6f-6687a2c29fdd"}'
Result:
{"errors":[{"message":"AiError: AiError: The model 214de035-8495-4c51-af6f-6687a2c29fdd does not exist. (01e3d7a2-3518-4e2d-a1a5-48b09dba46a0)","code":3043}]}
Happy to send you the AccounT, Finetune ID
I have already tried deleting and re-uploading the finetune via wrangler, which resulted in a new ID but the same error pattern. This confirms the issue is not related to a specific finetune asset.
• Failing Request ID: 01e3d7a2-3518-4e2d-a1a5-48b09dba46a0 (from the error message)Could you please have your engineering team investigate this internal discrepancy using the provided Request ID?
1. Proof the Finetune Exists (Successful GET Request):
This curl command, using my API token, correctly lists the finetune, proving it exists and my credentials are valid.
curl -X GET "https://api.cloudflare.com/client/v4/accounts/ACCOUNT_ID/ai/finetunes" \
-H "Authorization: Bearer {token}
Result:
{"success":true,"result":[{"id":"214de035-8495-4c51-af6f-6687a2c29fdd","name":"my-gemma-finetune", ...}]}
2,2. Proof the Inference Call Fails (Failing POST Request):
This curl command, using the exact same credentials and the ID from the successful call above, fails.
curl -X POST "https://api.cloudflare.com/client/v4/accounts/{account_id}/ai/run/@cf/google/gemma-3-12b-it" \
-H "Authorization: Bearer {token} \
-d '{"messages":[{"role":"user","content":"..."}],"lora":"214de035-8495-4c51-af6f-6687a2c29fdd"}'
Result:
{"errors":[{"message":"AiError: AiError: The model 214de035-8495-4c51-af6f-6687a2c29fdd does not exist. (01e3d7a2-3518-4e2d-a1a5-48b09dba46a0)","code":3043}]}
Happy to send you the AccounT, Finetune ID
I have already tried deleting and re-uploading the finetune via wrangler, which resulted in a new ID but the same error pattern. This confirms the issue is not related to a specific finetune asset.
• Failing Request ID: 01e3d7a2-3518-4e2d-a1a5-48b09dba46a0 (from the error message)Could you please have your engineering team investigate this internal discrepancy using the provided Request ID?
Welcome to the official Cloudflare Developers server. Here you can ask for help and stay updated with the latest news
86,942Members
Resources
Was this page helpful?
