Alniyat5f/WuQi-V3-Onyx-27B

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 17, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Alniyat5f/WuQi-V3-Onyx-27B is a 27 billion parameter Agentic language model developed by the WuQi project, fine-tuned from Qwen3.8-27B with a 32768 token context length. It is optimized for agentic tasks, demonstrating enhanced performance in complex multi-stage scenarios and significantly reduced token expenditure for reasoning. This model excels in planning, execution, and tool-use within agent environments, making it suitable for sophisticated automated workflows.

Loading preview...

WuQi-V3-Onyx-27B: An Efficient Agentic LLM

WuQi-V3-Onyx-27B is a 27 billion parameter Agentic model developed by the WuQi project, built upon the Qwen3.8-27B base model. It leverages LoRA for post-training on specialized domain data, resulting in a highly efficient model for agentic applications. The model maintains the native multimodal capabilities of its Qwen3.8-27B base and supports a substantial context length of 32768 tokens.

Key Capabilities & Optimizations

  • Agentic Performance: Specifically adapted for the DeepSeek Harness (DSH) agent environment, covering 12 types of tools (e.g., bash, web_search, run_code, subagent).
  • Long-Term Task Completion: Trained on multi-stage agentic trajectories (search, planning, coding, sub-agents, backend, verification), leading to significantly improved completion rates for complex, long-duration tasks compared to its base model.
  • CoT Optimization: Features optimized Chain-of-Thought (CoT) reasoning, resulting in an 86% reduction in 'Reason' token expenditure and a 60% decrease in total 'Output' token expenditure, enhancing efficiency.
  • Benchmark Performance: Achieves a total score of 2310 (96.2%) on the DSH-ABench-Basic benchmark, outperforming Qwen3.8-27B (93.8%) and approaching DeepSeekV4Flash (98.8%) on hard and comprehensive tasks.

Ideal Use Cases

  • Automated Agent Workflows: Excellent for applications requiring complex planning, tool invocation, and multi-step task execution.
  • Efficient Reasoning: Suitable for scenarios where minimizing token usage for reasoning and output is critical.
  • DSH-Compatible Environments: Optimized for integration into DeepSeek Harness-like agent frameworks.

Finetunes
2 models