odytrice/kenichi-flash
Kenichi Flash is a 24.1 billion parameter causal language model developed by odytrice, fine-tuned from Devstral Small 2 24B with a 32768 token context length. It is specifically optimized for domain-specialized code generation in F#, .NET, Svelte 5, TypeScript, Docker, and Kubernetes. This model excels as a fast agentic coding assistant, with all F# training samples verified by the F# compiler, making it ideal for targeted software development workflows.
Loading preview...
Kenichi Flash: Domain-Specialized Coding Assistant
Kenichi Flash is a 24.1 billion parameter coding model developed by odytrice, fine-tuned from Devstral Small 2 24B. It is uniquely specialized for fast, agentic code generation across specific technology domains, distinguishing it from general-purpose LLMs. The model was created using multi-teacher distillation from five frontier models, with all F# training samples rigorously verified by the F# compiler.
Key Capabilities
- Domain-Specific Code Generation: Highly proficient in F#, .NET/ASP.NET, Svelte 5/SvelteKit, TypeScript, Docker, and Kubernetes.
- Agentic Workflows: Optimized for tool use, multi-step reasoning, code review, and debugging within its specialized domains.
- High-Quality F# Output: F# code samples in the training data were compiler-verified, ensuring high accuracy and idiomatic style.
- Efficient Training: Trained in 1 hour 44 minutes on a single NVIDIA A100 GPU using Unsloth, demonstrating efficient resource utilization.
Good for
- AI Coding Assistants: Integrating into IDE plugins or standalone coding assistants for the specified technologies.
- Agentic Coding Pipelines: Automating multi-step coding tasks, code review, and refactoring.
- Specialized Development: Developers working extensively with F#, .NET, Svelte 5, TypeScript, Docker, or Kubernetes who need a highly focused coding partner.
- Rapid Prototyping: Generating clean, well-structured code quickly within its areas of expertise.