reaperdoesntknow/Qwen3-1.7B-Coder-Distilled-SFT

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Mar 25, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

reaperdoesntknow/Qwen3-1.7B-Coder-Distilled-SFT is a 1.7 billion parameter Qwen3-based causal language model developed by Convergent Intelligence LLC. It was created through a two-stage distillation process, first from a 30B Coder teacher for STEM reasoning, then fine-tuned on logical inference problems. This model excels at formal reasoning, propositional logic, and structured STEM derivation, making it suitable for tasks requiring explicit sequential logic and compositional reasoning.

Loading preview...

Model Overview

This model, reaperdoesntknow/Qwen3-1.7B-Coder-Distilled-SFT, is a 1.7 billion parameter Qwen3-based language model developed by Convergent Intelligence LLC. It is designed for structured reasoning and logical inference, built upon a unique two-stage training pipeline.

Key Capabilities & Training

  • Stage 1: Coder Teacher Knowledge Distillation: The model was initially distilled from a Qwen3-Coder-30B-A3B-Instruct teacher. This stage focused on transferring the teacher's method of organizing reasoning, emphasizing precise sequential logic, explicit state tracking, and compositional decomposition, particularly for STEM chain-of-thought problems across 12 domains.
  • Stage 2: Logical Inference SFT: Following distillation, the model underwent supervised fine-tuning on approximately 54,600 logical inference problems from the KonstantinDob/logic_inference_dataset. This stage activated the latent logical structure gained from the Coder teacher, making it explicit for formal propositional logic and logical entailment.
  • Mathematical Foundations: The training pipeline is grounded in Discrepancy Calculus, a measure-theoretic framework for preserving structural information during distillation.
  • Context Length: The model was trained with a context length of 1024 tokens.

Intended Uses

This model is particularly good for:

  • Logical inference and propositional logic
  • Formal reasoning and structured argumentation
  • STEM derivation and educational tutoring
  • Component in verification pipelines
  • Edge deployment via GGUF (quantized versions available)

It is not intended for general code generation, formal proof verification (e.g., Lean/Coq), safety-critical analysis, or tasks requiring long contexts beyond 1024 tokens.

Finetunes
2 models