Gemini Decisions API launches with input-only billing
Google’s November 11 release evaluates up to 1,000 independent questions per request; latency claims have narrower test conditions.
Google launched Gemini Decisions API on November 11, 2026. Gemini Decision 1 returns probabilities and confidence for yes/no, choice and rubric questions across text, images, audio, video and PDFs.
Requests support up to 256K input tokens and 1,000 independent questions. Published pricing is US$0.10 per million input tokens, with no output charges. Shared input is billed once per request; question instructions and criteria add input tokens. A no-cost prototyping tier exists, but its quotas were not verified.
Google’s 150–250 ms latency discussion covers contexts up to 16K tokens and 1–100 questions, not guaranteed performance at maximum capacity. Documentation warns that choices receive probabilities even when none fits: provide an exception option and human review.
Access is through Gemini API/AI Studio. Its general region list includes Taiwan, not Hong Kong or Macau; individual accounts were not tested. Chinese and Cantonese performance is not established by these sources. No hands-on testing was performed.
References:
https://aistudio.google.com/docs/decisions-api
https://aistudio.google.com/docs/available-regions
Requests support up to 256K input tokens and 1,000 independent questions. Published pricing is US$0.10 per million input tokens, with no output charges. Shared input is billed once per request; question instructions and criteria add input tokens. A no-cost prototyping tier exists, but its quotas were not verified.
Google’s 150–250 ms latency discussion covers contexts up to 16K tokens and 1–100 questions, not guaranteed performance at maximum capacity. Documentation warns that choices receive probabilities even when none fits: provide an exception option and human review.
Access is through Gemini API/AI Studio. Its general region list includes Taiwan, not Hong Kong or Macau; individual accounts were not tested. Chinese and Cantonese performance is not established by these sources. No hands-on testing was performed.
References:
https://aistudio.google.com/docs/decisions-api
https://aistudio.google.com/docs/available-regions
Usage conditions
- Hong Kong
- Hong Kong is absent from the general AI Studio/Gemini API list; alternative platforms were not verified.
- Macau
- Macau is absent from the general AI Studio/Gemini API list; alternative platforms were not verified.
- Taiwan
- Taiwan appears in the general service list; new-endpoint account eligibility and quotas were not tested.
Official source: Gemini Decisions API: Fast, calibrated multimodal decisions