Skip to content

Clef-Flash

Tier 1: 4.75 points · 1,000 tokens (500 input + 500 output)

Comparison method and assumptions.

Pricing

Website Pricing

Type Cost
Input Text 2.5 points/1k tokens
Output (Text) $0.20/1M tokens · 7 points/1k tokens
Input (File) $0.00000416/KB (64 est. tokens) · 0.16 points/KB

All parsed rates (both currencies):

Service Currency Amount Per
Input (text) USD 0.065 1000000 tokens
Input (text) POINTS 2.5 1000 tokens
Output (text) USD 0.20 1000000 tokens
Output (text) POINTS 7 1000 tokens
Input (file) USD 0.00000416 1 kb
Input (file) POINTS 0.16 1 kb

Last Checked: 2026-10-22 13:00:01.907691

Bot Information

Creator: @OpenSourceLab

Description: Clef-flash evaluates your input state alongside a schema of typed questions and returns a calibrated probability for every allowed option. This ultra-low-latency, open-weight decision model serves as a drop-in, Jev alternative tailored for high-volume, real-time pipelines. It supports PDF and XLSX files. Clef-Flash does not support video and audio attachments.

Developed by Cloudflare and released on October 1, 2026, Clef-flash is a 9-billion parameter model built on a post-trained Qwen3.5-9B backbone. Unlike traditional language models that sequentially generate free-form sentences, it processes requests via a non-autoregressive, prefill-only forward pass, scoring permitted choices in parallel through a joint decision head.

Extra: Powered by a server managed by @OpenSourceLab. Learn more

Architecture

Input Modalities: Unknown

Output Modalities: Unknown

Modality: unknown

Technical Details

Model ID: Clef-Flash

Object Type: bot

Created: Unknown

Owned By: OpenSourceLab

Root: Clef-Flash

API Last Updated: Not listed in API