Gemini Decisions API scores multimodal questions in 150-250ms with zero output tokens
McDonaghMatthew · x · 2026-11-18
Google's new Gemini Decisions API evaluates closed-set questions over PDFs, audio, video, images and text in a single forward pass — 150-250ms latency, zero output tokens, no separate OCR or transcription step, up to 256K input tokens.
It mixes three question types per call:
- Yes/no checks (e.g. does an invoice have a PO number, does a call include consent)
- Multiple choice for routing calls, documents or prompts across queues, tools and models
- Rubric rating of audio delivery, video events or document quality
The thesis: simple decisions don't need trillions of parameters generating open-ended text — calibrated structured answers suffice.
More from Models
- Polymarket opens book on when OpenAI reports another AI sandbox escape — Polymarket · 2026-11-18
- OpenAI ships 6.2 Astra and Sol, immediately panned as its worst model release — Hesamation · 2026-11-18
- GPT-6.2 draws backlash on X, but early hands-on feels faster even at compaction — CtrlAltDwayne · 2026-11-18
- ChatGPT 6.2 models panned: "vibes are bad off the charts" — minchoi · 2026-11-18
- Australian couple built a 220B sovereign AI model on under $10M, ranked 43rd worldwide — aakashgupta · 2026-11-18
- Users say Opus 5.5 refuses near-impossible tasks while rivals at least try — imjustnewatai · 2026-11-18