Cloudflare's Clef Brings Open Decision Models
The company pairs open-weight reasoning models with a reinforcement-learning platform aimed at operational decisions.
Cloudflare has introduced Clef, a family of open-weight "decision models" designed for reasoning tasks, released together with a new reinforcement-learning fine-tuning platform. The company detailed the launch on its engineering blog.
Unlike general-purpose chat models, Clef is framed around making decisions — the kind of structured, repeatable judgments that software systems make constantly. Pairing the models with an RL fine-tuning platform signals that Cloudflare expects users to adapt them to specific tasks rather than treat them as fixed off-the-shelf tools.
Why it matters
Open weights lower the barrier for teams that want to inspect, host, and specialize a model themselves. The bundled fine-tuning platform is the more interesting half of the story: reinforcement learning is how many of today's reasoning systems are sharpened, and making that workflow accessible could let smaller teams tune decision logic for their own domains.
- Open-weight models focused on reasoning and decision-making
- A reinforcement-learning platform for task-specific fine-tuning
- Released under a custom license, so terms are worth reviewing before production use
Few technical specifics — parameter counts, context length, or benchmark figures — accompanied the initial announcement, so the practical ceiling for Clef remains to be seen. For now, it marks Cloudflare's bid to make tunable decision models a building block developers can reach for directly.
Sources
More from NVIDIA
All NVIDIA releases →NVIDIA's Cosmos 3 Brings Open Interactive World Models
The new open release pushes beyond passive video generation toward simulated worlds that respond to input.
NVIDIA Releases Cosmos 3 for World Simulation
The new open models push NVIDIA's world-foundation effort toward interactive, real-time video generation.

NVIDIA releases compact Dory 1.0 4B text model
The new Nemotron Labs model packs a 4-billion-parameter language model trained on NVIDIA's Nemotron datasets.
More in Text / LLM
All Text / LLM →
Microsoft's Aion 1.0 Instruct is a compact 2.3B chat model
A small, multilingual instruction-tuned model aimed at chat and summarization lands on Hugging Face.
SenseTime Releases Looped 9B Multimodal Model
The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Tencent releases Hunyuan Hy4-flash multimodal MoE
The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.
0 comments
No comments yet. Be the first to weigh in.