CodeXomics/CodeXomics-ToolAgent-4B-v1

VISIONConcurrent Unit Cost:1Model Size:4.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 10, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

CodeXomics-ToolAgent-4B-v1 is a 4.2 billion parameter tool-calling model fine-tuned by CodeXomics from Qwen/Qwen3.5-4B. It is specifically optimized for native function calling within the CodeXomics genomics workbench, handling tasks like sequence analysis, annotation, and database retrieval. This model demonstrates 100% accuracy on the CodeXomics Benchmark, significantly outperforming its base model for genomic workflows. It is designed for integration into the CodeXomics AI-native genome browser for driving biological analyses.

Loading preview...

CodeXomics-ToolAgent-4B-v1: Specialized Genomics Tool Agent

CodeXomics-ToolAgent-4B-v1 is a 4.2 billion parameter model, fine-tuned from Qwen/Qwen3.5-4B, specifically designed for advanced tool-calling within the CodeXomics genomics workbench. This model excels at interpreting natural language commands and translating them into native function calls for a wide array of genomic tasks.

Key Capabilities & Features

  • Native Tool Calling: Performs function calls against the CodeXomics tool registry, covering file operations, sequence analysis, annotation, track control, export, BLAST, primer design, and UI control.
  • High Accuracy: Achieved 100% accuracy on the CodeXomics Benchmark (172 tests), a significant improvement over the base Qwen3.5-4B model (95.9%).
  • Optimized Performance: Demonstrates reduced average latency per test (10.7s vs. 12.0s for base model) and per tool call (7.3s vs. 7.9s).
  • Fine-tuning Method: Utilizes QLoRA with a maximum sequence length of 3,072 tokens, trained on the CodeXomics-ToolCalling-v1 dataset.

Ideal Use Cases

  • AI-Native Genome Browsers: Specifically built for the CodeXomics desktop application to drive genome visualization and biological analyses via conversational AI.
  • Genomic Workflow Automation: Automating complex genomic tasks through natural language interaction.
  • Tool-Augmented Research: Enabling researchers to interact with genomic tools and databases more intuitively.

This model is highly specialized for genomic workflows and its generalization to other tool-calling domains has not been evaluated.

Finetunes
1 models