Google accidentally leaks gemini-decision-1: 262K context, free output, tested live then pulled
AGI Hunt · wechat · 2026-11-12
A developer stumbled upon an unreleased Google model, gemini-decision-1, in the model list: 262K context, $0.1 per million input tokens, free output.\n\nIt is a "decision model" — no chatting, only three types of judgmental answers: yes/no, which one, and a score, each with probabilities, suited for classification, routing and scoring. The author got it working with his own key: endpoint POST v1beta/decisions with x-goog-api-key auth; the format differs from Gemini's generateContent but closely matches OpenAI's Decisions API, even the error messages. It supports text, image, document, audio and video inputs.\n\nIn a live test on a support ticket ("charged $49 twice, will dispute with bank if no refund"), the model returned: refund request (0.99), billing queue (0.995), highest urgency (0.84). Soon after, the model stopped working — returning model not found — and disappeared from Google's model list (63 → 62), suggesting an accidental leak quickly rolled back. Jev, OpenAI, Microsoft and Qwen have already shipped decision models (Microsoft's reportedly trained on Qwen).
More from Models
- Open-weights Inkling Small 2 impresses: researcher says he prefers it over DeepSeek V4.1 Flash — gnukeith · 2026-11-13
- Manual Review Exposes How LLMs Cheat Benchmarks: Sandbox Escapes and Copying Peers — teortaxesTex · 2026-11-13
- Models invent their own filler tokens during RL training — code_star · 2026-11-13
- Google Search reportedly gets massive Gemini 4 Hydrogen upgrade with in-search simulations — AI_Andrew · 2026-11-13
- Chinese medical AI beats 250 top doctors 88-12 on complex lung cancer cases in new study — 量子位 · 2026-11-13
- Gemini 4 Carbon ties Argon at 93/98 but burns 2.67× output tokens and 2.22× Python runs — Correct_Tomato1871 · 2026-11-13