Cloudflare Clef and Clef-Flash
Cloudflare's first in-house trained models: open-weight decision models (27B and 9B) that return calibrated probabilities instead of generated text for agent routing, classification, and guardrails.
Why it matters
The first major infra vendor to ship open-weight decision models under Apache 2.0 — a Jev-compatible API with vision input and 38.8 ms median decision latency, turning bounded agent decisions into a commodity layer.
What you could build with it
A support-ops startup could put Clef-flash in front of every incoming ticket: one call scores urgency and picks the owning team from a fixed list, auto-routing high-confidence cases and escalating only the uncertain band to humans — cutting triage latency from seconds to milliseconds.
Does it hold up?
Usable today via Workers AI, Hugging Face weights, Ollama, and community MLX builds; vendor benchmarks are self-reported and one independent 42-case test found Clef-flash trailing Jev on nuanced triage, so test on your own workload before trusting it.
Built with Cloudflare Clef and Clef-Flash
- Cloudflare Clef: open-weight decision models on Workers AIdatanorth.ai · Benchmark and latency comparison against Jev; notes four decision models landed within eight days, signaling the category is becoming infrastructure.
- Cloudflare Clef on Mac: Flash fits 16 GB, 27B needs 32 GBmodelfit.io · Hands-on local-run guide: community 4-bit MLX builds run Clef-Flash on a 16 GB Mac, while GGUF quants lack the joint schema head and cannot score decisions yet.
- Cloudflare Clef targets 38.8ms AI agent decisionsbitcoinversus.tech · Launch coverage with a video walkthrough testing the 27B Clef locally across text, image, and video decision tasks.
Learn more
First spotted on hf: source.
More AI models
ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71Mistral Forge: enterprise platform for training and continuously improving proprietary modelsMistral AI launched Forge (Sep 28) — a full-lifecycle model training platform (pre-training, SFT, DPO/ODPO, RL…models · JEV 0.69DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69kev: open, trainable Jev-like family of small decision models on Qwen3.5/3.8A family of small decision models (0.8B to 27B) built on Qwen3.5/3.8 that reproduces Jev's prefill-only typed-decision…models · JEV 0.68GPT-Synopsys: OpenAI and Synopsys multi-year deal to build an AI model that operates EDA toolsOpenAI and Synopsys signed a multi-year agreement on Sep 30 to jointly develop GPT-Synopsys, a specialized model that…models · JEV 0.66
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.