Nov 11, 2026 — Google enters the Decision Model market. At AI Engineer, Cloudflare shows what parts of a build failure a classifier can handle. TypeSafe AI showcases their production customers.
Nov 12 · 24 min read
Today is “System One” day at the 2nd Annual AIE CODE summit, and everything is fitting together so serendipitously:
-
Cloudflare’s Michelle Chen broke down their Clef and Clairic models
-
Richard covered Diogo’s triumphant return to AIE to talk through the “Jevolution” of their most representative production customers since crossing $100M ARR
-
In a huge surprise, Google beat Anthropic to a production decision model, with Sundar and Logan announcing Gemini Decision 1, a calibrated multimodal decision model supporting up to 1000 questions at $0.10/M input tokens and ~200ms e2e latency.

Benchmarks look good (usual benchmark disclaimers apply):

You can hear all of this and more on today’s System One track.
Support AINews by setting up your Agents with SerpApi to read and scrape the web!
Sponsored by SerpApi
Give your AI agents full pages, not just links
Retrieve public web pages by URL as clean Markdown, structured JSON, or raw HTML with SerpApi’s Extract API. Extract a single page, map a site’s URLs, or run a scoped same-site crawl. Just set engine=extract with a URL and add output=md for Markdown output. Pair it with 100+ SerpApi search APIs to go from query to page content through a single provider.
Thanks to SerpApi for partnering with business@latent.space!
AI Twitter Recap
Google Launches Gemini Decision 1 for Single-Pass Classification
-
Gemini Decisions API: Google launched a new API powered by Gemini Decision 1, a model built for closed-set decisions so it can score choices in a single forward pass instead of generating text token by token (Sundar Pichai, Logan Kilpatrick, Google).
-
Specs: Supports yes/no, multiple choice, and rubric scoring with up to 255 choices for multiple choice. It accepts text, image, audio, video, and PDF inputs, handles up to 1,000 questions per request, has a 256K context window, returns answers in 150–250ms end-to-end, and includes a confidence score with every answer (@_philschmid).
-
Pricing: $0.10 per 1M input tokens and $0 for output. Google says that makes it roughly 5x cheaper than Flash-Lite (Sundar).
-
Origin: The project started as an internal prototype at Google DeepMind, and a small team shipped the model and API roughly six weeks later (@_philschmid).
-
-
Google vs. TypeSafe’s Jev: Google’s own testing shows Decision 1 beating Jev 1.14 on 5 of 7 shared benchmarks, including 92.2% vs. 90.0% on JevBench. Jev is priced at $0.042 per 1M input tokens (@kimmonismus).
-
Artificial Analysis Evaluation: Independent evaluations were more mixed (Artificial Analysis).