FRPO/qwen3-1.7b-a14_shuffle-k1-cNone-shuf-clip0.2-mb4-eta100-bs256x5-n2

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 15, 2026Architecture:Transformer Featherless Exclusive Cold

FRPO/qwen3-1.7b-a14_shuffle-k1-cNone-shuf-clip0.2-mb4-eta100-bs256x5-n2 is a 2 billion parameter language model based on the Qwen3-1.7B architecture, developed by FRPO. This checkpoint is a result of Reinforcement Learning (RL) fine-tuning experiments using the KL-in-LLM-RL / FRPO methodology. It is specifically designed for research and experimentation in RL-based language model optimization, offering a 32768 token context length.

Loading preview...

Model Overview

FRPO/qwen3-1.7b-a14_shuffle-k1-cNone-shuf-clip0.2-mb4-eta100-bs256x5-n2 is a 2 billion parameter language model derived from the Qwen/Qwen3-1.7B base model. It represents a specific checkpoint from the KL-in-LLM-RL / FRPO experimental series, fine-tuned using Reinforcement Learning (RL) techniques with the verl framework.

Key Characteristics

  • Base Architecture: Qwen3-1.7B, a 2 billion parameter model.
  • Fine-tuning Method: Reinforcement Learning (RL) using the KL-in-LLM-RL / FRPO experimental setup.
  • Checkpoint: global_step_200 is provided in fp32 safetensors format, directly as saved by the trainer.
  • Context Length: Supports a context window of 32768 tokens.
  • Experimental Focus: This model is a product of research into RL-based optimization for large language models.

Intended Use Cases

This model is primarily suited for:

  • RL Research: Experimenting with and evaluating the effects of KL-in-LLM-RL fine-tuning on language model performance.
  • Comparative Analysis: Studying how specific RL configurations (encoded in the repo name, e.g., a14_shuffle-k1-cNone-shuf-clip0.2-mb4-eta100-bs256x5-n2) impact model behavior and capabilities.
  • Developer Tooling: As a foundation for further fine-tuning or integration into applications requiring a Qwen3-1.7B variant with specific RL-tuned characteristics.

Finetunes
3 models