bobber
/

routangseng-0.8b-hottake

political-analysis

Model card Files Files and versions

肉糖生 0.8B Hot-Take

0.8B model distilled from the 4B Phase 10 Think-SFT teacher. Designed as a sharp, quick-take machine for mobile/WebGPU deployment.

Model Info

Base: huihui-ai/Huihui-Qwen3.5-0.8B-abliterated
Teacher: bobber/routangseng-phase10-think-sft (4B)
Training: SFT on 949 condensed hot-take examples (200-400 chars)
Eval: Heuristic score 4.60/5

ONNX / WebGPU

ONNX model: bobber/routangseng-0.8b-hottake-onnx
GitHub CDN: bobbercheng/routangseng-models
WebGPU demo: bobber/routangseng-chat

Production Note

Strip </think> tags at inference time — the model occasionally emits them as artifacts.

Related

GPU Space: bobber/routangseng-chat-gpu
4B recommended: bobber/routangseng-phase10-think-sft
Project docs: bobber/routangseng-qwen35-4b-project

Downloads last month: 8

Safetensors

Model size

0.8B params

Tensor type

BF16

·

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for bobber/routangseng-0.8b-hottake

Base model

Qwen/Qwen3.5-0.8B-Base

Finetuned

Qwen/Qwen3.5-0.8B

Finetuned

huihui-ai/Huihui-Qwen3.5-0.8B-abliterated

Finetuned

(7)

this model

Quantizations