If you get a "this model is not available for your subscription tier" on the /v1/messages endpoint when trying to use Digital Ocean's new Serverless Inference, it's likely because of a fine-print gotcha buried in the docs (https://docs.digitalocean.com/products/inference/details/limits/)
> All tiers, except Tier 1 and Tier 2, have access to all models. Tier 1 and Tier 2 do not have access to any Anthropic models and OpenAI models (except gpt-oss-120b and gpt-oss-20b).