WithinUsAI/Qwen3-rStar.Coder.Expert-IF-0.6B

TEXT GENERATIONConcurrent Unit Cost:1Model Size:0.8BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Feb 1, 2026Architecture:Transformer Featherless Exclusive Cold

WithinUsAI/Qwen3-rStar.Coder.Expert-IF-0.6B is a 0.8 billion parameter Qwen3-based decoder-only transformer model, created by WithinUsAI through a SLERP merge of rStar-Coder and IF-Expert variants. This model specializes in high-skill code generation, particularly for Python and TypeScript, combined with precise instruction following. It excels at producing instruction-compliant code, making it suitable for complex coding tasks requiring strict adherence to multiple constraints.

Loading preview...

Model Overview

WithinUsAI/Qwen3-rStar.Coder.Expert-IF-0.6B is a 0.8 billion parameter model built on the Qwen3 architecture, developed by WithinUsAI. It was created using a SLERP (Spherical Linear Interpolation) merge technique, combining two specialized base models:

  • prithivmlmods/rStar-Coder-Qwen3-0.6B: Contributes expert code generation capabilities.
  • suayptalha/Qwen3-0.6B-IF-Expert: Provides precise instruction following and fidelity.

This strategic merge aims to leverage the strengths of both components, resulting in a model that can generate production-grade code while strictly adhering to complex instructions and formatting requirements. The merge configuration involved specific weighting for different model components, such as 100% rStar-Coder embeddings and 100% IF-Expert output head, with a 50/50 interpolation for self-attention and MLP layers.

Key Capabilities

  • Expert Code Generation: Excels in generating Python and TypeScript code with algorithmic precision.
  • Precise Instruction Following: Demonstrates strong adherence to complex instructions and format constraints.
  • Instruction-Compliant Code: Produces code that reliably satisfies multiple constraints specified in a single request, surpassing the capabilities of its individual base models.

Ideal Use Cases

This model is particularly well-suited for scenarios requiring:

  • Generating code that must meet multiple, specific requirements (e.g., a FastAPI endpoint accepting CSVs, validating data, returning paginated JSON, and including OpenAPI docs).
  • Tasks where strict adherence to complex instructions and output formats is critical for code functionality.
  • Development of applications needing reliable, constraint-driven code generation.

Finetunes
1 models