Clef

Overview

Clef is a family of open-weight decision models announced October 2, 2026 by Cloudflare — Clef (27B) and Clef-flash (9B), licensed under Apache 2.0 and hosted on Hugging Face. Clef is aimed at classification, routing, and guardrail workloads rather than open-ended generation, and ships alongside a new RL fine-tuning platform wiring together AI Gateway, Workers AI, and Trainer so tuning + serving + routing live in one developer-plane surface. Clef is a control-plane entry, positioned to sit in front of frontier LLMs like Claude Opus 5.5 or Gemini 4 Argon, not a frontier-generation competitor.

Timeline

• 2026-10-02-AI-Digest — Cloudflare releases Clef (27B) and Clef-flash (9B) — open-weight decision models under Apache 2.0 on Hugging Face, aimed at classification, routing, and guardrail workloads rather than open-ended generation. Alongside the weights, Cloudflare shipped a new RL fine-tuning platform wiring together AI Gateway, Workers AI, and Trainer, so tuning + serving + routing live in one developer-plane surface. Load-bearing softener: this is a control-plane entry, not a frontier-generation competitor — Clef is sized and positioned for the guardrail / classification tier that sits in front of LLMs like Claude Opus 5.5 or Gemini 4 Argon. Monetisation is implicit via Workers AI pricing on inference and fine-tuning compute; no revenue guide published. Reframe worth carrying: hyperscaler-backed open-weight entry at the control-plane layer, with the serve + tune pipeline bundled on Workers AI, not Cloudflare launches a frontier LLM. Clef is the second hyperscaler-attached open-weight release at the decision-model tier this quarter — the pattern of “open at the control plane, closed at the frontier” is consolidating. Log against MOC - Open Source Models and MOC - Developer Tools.

Key Developments

1. 27B + 9B Open-Weight Decision Models Under Apache 2.0 — Control-Plane Entry, Not Frontier-Generation Competitor (October 2, 2026): Cloudflare‘s Clef and Clef-flash land on Hugging Face as Apache 2.0 open weights sized for classification, routing, and guardrail workloads — the tier that sits in front of frontier LLMs rather than competing with them on open-ended generation. Load-bearing framing to carry: the “open at the control plane, closed at the frontier” pattern is consolidating — this is the second hyperscaler-attached open-weight release at the decision-model tier this quarter. Clef is the model layer of a stack whose fine-tuning + serving + routing pipeline ships bundled on Workers AI; monetisation is implicit via Workers AI pricing, with no revenue guide published. 30/60/90-day watch: whether independent benchmarks surface Clef vs peer control-plane models (Mistral Shieldstral, Granite Guardian); whether Clef adoption shows up in Cloudflare’s Q4 Workers AI metrics; whether a third hyperscaler-attached open-weight decision model lands inside the quarter.

Related

See also: Cloudflare, Claude Opus 5.5, Gemini 4 Argon, MOC - Open Source Models, MOC - Developer Tools.