I think AI companies should throttle requests and downgrade to a cheaper model when inference is used up. Just like phone companies do when you used up 5g, but still can send texts
@mac_eth Better. Do we really need to list an offering per model? I’d rather just post an api key with a set discount, not per model basis. It’s much easier to handle
@mac_eth Hey, theres a naming issue with the openai models when trying to add with venice key
error: "HTTP 404: {\"error\":\"Specified model not found: gpt-5.3-codex. Did you mean: openai-gpt-52-codex, openai-gpt-53-codex, openai-gpt-oss-120b?\"}"
Same with gpt 5.4 / 5.5