Tesslate/OmniCoder-9B
Tesslate/OmniCoder-9B is a 9-billion parameter coding agent model, fine-tuned on Qwen3.5-9B's hybrid architecture. It was trained on over 425,000 curated agentic coding trajectories, including real-world software engineering tasks and tool use. This model excels at complex code generation, error recovery, and multi-step reasoning, making it highly effective for developer-centric applications.
Loading preview...
OmniCoder-9B: A 9B Coding Agent
OmniCoder-9B, developed by Tesslate, is a 9-billion parameter coding agent model built upon the Qwen3.5-9B's hybrid architecture, which features Gated Delta Networks interleaved with standard attention for efficient long-context processing. It was fine-tuned on an extensive dataset of over 425,000 curated agentic coding trajectories, derived from frontier models like Claude Opus 4.6, GPT-5.4, GPT-5.3-Codex, and Gemini 3.1 Pro.
Key Capabilities
- Advanced Agentic Behavior: Learns from real-world agent trajectories, demonstrating strong error recovery (read-before-write), responsiveness to LSP diagnostics, and minimal edit diffs.
- Hybrid Architecture: Inherits Qwen3.5's architecture, supporting a 262K native context window, extensible to over 1M tokens.
- Complex Reasoning: Supports
<think>...</think>reasoning chains for effective problem decomposition. - Strong Performance: Achieves 83.8% on GPQA Diamond (pass@1), 90% on AIME 2025 (pass@5), and 23.6% on Terminal-Bench 2.0, showing significant improvements over its base model.
Good For
- Automated Code Generation: Ideal for generating complex code, handling multi-step programming tasks, and integrating with developer tools.
- Agentic Workflows: Suitable for applications requiring intelligent agents that can recover from errors and perform sophisticated reasoning.
- Software Engineering Tasks: Excels in scenarios involving real-world software development, tool use, and terminal operations.