ermiaazarkhalili/FastContext-4B-SFT_base-SFT-Claude-Opus-Reasoning-Unsloth

TEXT GENERATIONConcurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 20, 2026Architecture:Transformer0.0K Featherless Exclusive Cold

The ermiaazarkhalili/FastContext-4B-SFT_base-SFT-Claude-Opus-Reasoning-Unsloth is a 4 billion parameter language model, fine-tuned from Microsoft's FastContext-1.0-4B-SFT using LoRA and Unsloth. This model is specifically supervised fine-tuned on a private Claude reasoning distillation dataset, indicating an optimization for reasoning tasks. It utilizes the Qwen3ForCausalLM architecture and has a context length of 32768 tokens, making it suitable for applications requiring strong reasoning capabilities within a substantial context.

Loading preview...

Model Overview

This model, ermiaazarkhalili/FastContext-4B-SFT_base-SFT-Claude-Opus-Reasoning-Unsloth, is a 4 billion parameter language model built upon Microsoft's FastContext-1.0-4B-SFT base model. It leverages the Qwen3ForCausalLM architecture and has been fine-tuned using LoRA (Low-Rank Adaptation) via Unsloth and TRL.

Key Capabilities & Training

  • Reasoning Optimization: The model was supervised fine-tuned on a private dataset, ermiaazarkhalili/claude-reasoning-distillation, specifically targeting reasoning tasks. This suggests an enhanced ability to process and generate logical responses.
  • Efficient Fine-tuning: Utilizes LoRA with a rank of 16 and alpha of 16, trained for 1 epoch with a learning rate of 0.0002, demonstrating an efficient fine-tuning approach.
  • Context Length: Inherits a substantial context length of 32768 tokens, allowing it to handle extensive inputs for complex reasoning scenarios.

Limitations

  • No Benchmark Evaluation: Currently, no downstream benchmark evaluations have been performed on this specific checkpoint; only training loss observations are available.
  • Inherited Biases: The model inherits biases, knowledge cutoffs, and potential failure modes from its base model.
  • Specific Fine-tuning: Its behavior outside the distribution of the single instruction-following dataset it was fine-tuned on remains untested.
  • Merged Adapters: The LoRA adapters are merged into the base weights, meaning the fine-tune cannot be detached from the base model.