All news

Amazon and Cloudflare release open "decision models": fast SIs that choose an option rather than write text

Prepared by Sintelligent with the help of SI. Publication time restored from the build log. . Updated: .

A class of small, very fast SI "deciders" that can be run locally is emerging, making automation more accessible and cheaper.

AWS released Strands Decider 2B based on Qwen3.5-2B (Apache 2.0): according to AI Weekly, 72% on JevBench v19 and a median latency of about 106 ms on an RTX 3090. Cloudflare opened Clef and Clef-flash with a 64 thousand-token context and image support; by its own comparison, Clef's median latency is 209 ms versus 524 ms for Jev from TypeSafe AI. Both models respond with a specific choice from given options along with a confidence score, making them competitors to Jev and OpenAI's Decisions API.

History

Models that answer with a choice or a probability, not text

7 events
  1. OpenAI opens the Decisions API to all customers: the model quickly picks from given options, answering in 100–180 milliseconds instead of 1–3 seconds
  2. Subquadratic releases SubQ Snap, a decision model for agents that processes 100 thousand tokens in about a second
  3. Cloudflare releases open model Clef-omni that picks an answer from audio, video, images and text in one pass
  4. Microsoft releases Decision-1: a model that answers with the probability of a choice rather than text, priced at $0.042 per million input tokens
  5. Liquid AI releases open d1 models that answer without text, in a single pass
  6. OpenAI opens Decisions API: the model answers with a probability, a choice or a score instead of text
  7. Amazon and Cloudflare release open "decision models": fast SIs that choose an option rather than write textOpen now
Follow on Telegram