kavai/ qwen3.5-GPT5:0.8b-q8_0

954 3 days ago

Qwen 3.5 is a family of open-source models that delivers exceptional utility and performance for tool calling and Agentic abilities. Smaller Models may suffer from slower speeds.

vision tools thinking 0.8b 2b 4b 9b 27b 35b 122b 
ollama run kavai/qwen3.5-GPT5:0.8b-q8_0

Details

3 days ago

a69ef2815666 · 1.0GB ·

model
qwen35
·
873M
·
Q8_0
system
You are an expert AI programming assistant, working with a user in the VS Code editor. When asked fo
license
Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR US
params
{ "presence_penalty": 1.5, "temperature": 1, "top_k": 20, "top_p": 0.95 }

Readme

Qwen3.5-GPT5

Qwen3.5-GPT5 builds on the strong foundation of the Qwen3.5 family with a refined system prompt optimized for reasoning clarity, instruction adherence, and developer-centric workflows. The goal is to provide a drop-in upgrade that improves consistency, structured outputs, and general usefulness across coding, analysis, and conversational tasks. The model is has equal of if not better Tool calling capabilities as the base model.

Highlights

  • Improved reasoning stability and reduced hallucination
  • Stronger instruction following and formatting discipline
  • Better performance in coding and technical explanation tasks
  • Optimized for developer workflows and tool usage
  • Lightweight system-layer modification (no weight changes)

Model Variants

  • 122b — Maximum capability and reasoning performance
  • 35b — Balanced performance and efficiency
  • 9b — Fast and lightweight deployment

Usage

ollama run kavai/qwen3.5-GPT5:122b

or

ollama run kavai/qwen3.5-GPT5:35b

Design Philosophy

This release focuses on practical usability improvements without modifying base model weights. By injecting an optimized system layer, Qwen3.5-GPT5 enhances behavioral alignment while preserving the strong general intelligence of the original Qwen3.5 models.

Benchmarks

Formal benchmark results are coming soon. Model speeds may be slower than expected due to the system prompt.

Notes

  • No architectural changes to base model
  • Fully compatible with existing Qwen3.5 tooling
  • Suitable for chat, coding, and reasoning workload.
  • Recommended to use on GPU.

License

Follows the original Qwen3.5 model license.