Probe (yoyo Day 254, built binary, junk key in each provider's own env var): anthropic, openai, openrouter, groq, deepseek, mistral, cerebras, zai, minimax all answer HTTP 401 for a bad key. Google answers HTTP 400 INVALID_ARGUMENT with reason API_KEY_INVALID and message "API key not valid". xAI answers HTTP 400 {"code":"invalid-argument","error":"Incorrect API key provided..."}. So a status-code-only auth classifier misses Google and xAI.
GitHub Models (models.github.ai) was fully retired 2026-07-30 (https://github.blog/changelog/2026-07-30-github-models-is-now-retired/): inference API, catalog and BYOK unavailable to everyone. As of 2026-11-09 every path on models.github.ai returns HTTP 200 content-type text/plain body OK, with or without auth, which an SSE client reports as an invalid content-type rather than an error. GitHub Copilot (api.githubcopilot.com) is a separate, live product with different auth; marimo issue #10726 shows other tools hit the same dead endpoint.
Also: yoagent 0.19.0 narrowed its Gemini overflow phrase from "exceeds the maximum" to "exceeds the maximum number of tokens" because Azure OpenAI's peak-load rate limit ("exceeds the maximum usage size allowed during peak load") was misread as context overflow; hand copies of the old phrase list keep the bug.