{"slug": "gemini-decisions-api", "title": "Gemini Decisions API", "summary": "Google's Gemini Decision 1 decision model is priced at 10 cents per million input tokens with free output, matching OpenAI's GPT-6 Luna decision model and sitting above Jev at 4.2 cents per million, per Simon Willison's test of the API. Unlike Jev, which handles text and JSON only, and OpenAI's model, which handles text and images, Gemini Decision 1 accepts text, images, audio, video and PDFs. In Willison's video test, Gemini returned probability scores rather than the 1.0 or 0.0 yes/no answers Jev and OpenAI produce, giving 0.9234 for \"Does this video have a robot in it?\" and 0.0498 for \"Does this video have a dragon in it?\" on his Yonkers stomping robot video.", "body_md": "It's priced the same as the GPT-6 Luna decision model: 10 cents per million input tokens, and output is free. Jev remains the cheapest of the three, at 4.2 cents per million.\n\nJev is text (and JSON) only. OpenAI can handle text and images. Gemini Decision 1 is fully multi-model: it covers text, images, audio, video, and PDFs.\n\nI tested it against video by passing these questions to it along with my [Yonkers stomping robot](https://simonwillison.net/2026/Nov/10/yonkers-rises/) video file:\n\n```\ndragon: Does this video have a dragon in it?\nrobot: Does this video have a robot in it?\nfrance: Is this video set in France?\nusa: Is this video set in the USA?\n```\n\nJev and OpenAI both return 1.0 or 0.0 for yes/no questions, but Gemini insists on being a little less confident than that:\n\n```\n{\n  \"dragon\": {\n    \"type\": \"yes_no\",\n    \"probability\": 0.04976021841520312\n  },\n  \"robot\": {\n    \"type\": \"yes_no\",\n    \"probability\": 0.9234055387792485\n  },\n  \"france\": {\n    \"type\": \"yes_no\",\n    \"probability\": 0.08329171353166241\n  },\n  \"usa\": {\n    \"type\": \"yes_no\",\n    \"probability\": 0.9230459995157706\n  }\n}\n```\n\nVia [Hacker News](https://news.ycombinator.com/item?id=50456791)\n\nTags: [ai](https://simonwillison.net/tags/ai), [generative-ai](https://simonwillison.net/tags/generative-ai), [gemini](https://simonwillison.net/tags/gemini), [llm-release](https://simonwillison.net/tags/llm-release), [jev](https://simonwillison.net/tags/jev), [decision-models](https://simonwillison.net/tags/decision-models)", "url": "https://wpnews.pro/news/gemini-decisions-api", "canonical_source": "https://simonwillison.net/2026/Nov/11/gemini-decisions-api/", "published_at": "2026-11-11 22:03:25+00:00", "updated_at": "2026-11-11 22:22:26.970339+00:00", "lang": "en", "topics": ["artificial-intelligence", "generative-ai", "large-language-models", "ai-products", "ai-tools"], "entities": ["Google", "Gemini Decision 1", "OpenAI", "GPT-6 Luna", "Jev", "Simon Willison"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/gemini-decisions-api", "markdown": "https://wpnews.pro/news/gemini-decisions-api.md", "text": "https://wpnews.pro/news/gemini-decisions-api.txt", "jsonld": "https://wpnews.pro/news/gemini-decisions-api.jsonld"}}