Century Report arc
Decision Models as a Distributed Judgment Layer
Small, locally runnable AI models are turning bounded judgments into cheap software primitives, distributing the ability to route, rank, verify, and constrain automated actions while raising questions about whose criteria those judgments encode.
- Status
- Active
- Developments
- 8
- Last 30 days
- 6
- Last covered
Current synthesis
Where it stands
Open-weight decision models are proliferating fast: Jev reached 35% of Vercel's paid teams in two weeks (10-17), while Microsoft, Cloudflare, Liquid AI and Nace AI ship competing models (10-10) and Perplexity added a joint-request decider format for faster batched judgment (10-19). Much of the category still runs on Qwen foundations; the live question is whether rapid entry sustains differentiation as judgment becomes a cheaper, interchangeable software primitive.
Complete record
Development timeline
8 qualifying developments, newest first
October 2026
-
[SCAN] Liquid AI's open-weight LFM3 family includes d1-2B and d1-Omni-2B decision models, extending locally fine-tunable judgment primitives across text and multimodal inputs.
-
[SCAN+TCR] Perplexity released open weights for pplx-decider-v1.2, a 27B decision model answering multiple questions in one pass for up to 5x faster inference, plus a roughly 4B Flash version.
-
[SCAN+TCR] Jev reached 35% of Vercel AI Gateway's paid teams in two weeks, its fastest recorded model adoption and nearly twice Sonnet 5.5's comparable reach.
-
[DD+POD] TypeSafe raised $870 million at a $7.5 billion valuation 24 days after releasing its Jev decision model, as Microsoft and Cloudflare priced rival decision models below it on downloadable weights.
-
[DD] Microsoft released Decision-1 at $0.042 per million input tokens with free output, Cloudflare cut Clef-flash below rival Jev and added multimodal Clef-omni, and Liquid AI and Nace AI shipped competing open-weight decision models.
-
[DD+POD] A congestion-aware recommender deployed in NYC's 2025-26 high school match raised the share of applicants ranking a nearby high-performing program from 10.5% to 16.4% in a randomized trial, simulating applicant-pool-wide responses before issuing any suggestion.
-
[SCAN+TCR] Liquid AI added image input to its probability-scored d1 decision model, claiming parity or better against GPT-6.1 Sol on four of six applications at 19 to 200 times lower cost.
-
[DD+POD] Amazon released Apache-licensed Strands Decider 2B and Cloudflare released open Clef models alongside OpenAI's Decisions API, bringing low-cost typed choices to local and hosted software.