---
license: apache-2.0
base_model: Cloudflare/clef
pipeline_tag: text-generation
library_name: transformers
tags:
- bf16
- clef
- full-precision
- transformers
- vllm
- xinference
---

# clef-bf16 (bf16)

Full-precision (bf16) PyTorch port of [Cloudflare/clef](https://huggingface.co/Cloudflare/clef),
for vLLM / TGI / Xinference / transformers users who don't want the GGUF quant.

- Format: `pytorch_model.bin` (bf16). safetensors conversion in progress.
- Architecture: identical to the original — drop-in for `AutoModelForCausalLM.from_pretrained("rty813/clef-bf16")`.
- Also loads in vLLM (`vllm serve rty813/clef-bf16`) and Xinference (search "clef").

Original model card:

---
license: apache-2.0
library_name: transformers
pipeline_tag: image-text-to-text
base_model: Qwen/Qwen3.8-27B
base_model_relation: finetune
tags:
- clef
- cloudflare
- systemone
- decision-model
- qwen3.8
- post-train
- image-text-to-typed-output
- multimodal
- structured-output
- classification
- custom-code
---

# Clef

- **Announcement:** [Clef decision models on the Cloudflare blog](https://blog.cloudflare.com/clef-decision-models)
- **Decision Index leaderboard:** [clef-evals.workers-ai-mle.workers.dev](https://clef-evals.workers-ai-mle.workers.dev)

Clef is a 27B multimodal model that turns a state and a schema of typed
questions into decisions. It reads the state as text, JSON, images, or video, and returns a
probability for every allowed option of every question in a single forward pass. There is no
free-form text generation and no output parsing.

The Clef API is fully compatible with Jev and SystemOne.

Clef is post-trained from [Qwen/Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B). See
[Clef-Flash](https://huggingface.co/Cloudflare/clef-flash) for the
smaller, faster variant.

## Model

- **Backbone:** Qwen/Qwen3.8-27B with its vision encoder, stored as standard sharded safetensors.
- **Joint schema head:** a small transformer head that reads the backbone's final hidden states,
  routes evidence from the state to each question, and scores all options of all questions jointly.
- **Output:** one logit per allowed option for each question. Apply a softmax per