FRPO/qwen3-4b-a1_base-k1-cNone-clip0.2-mb4-eta100-bs256x5-n4

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 15, 2026Architecture:Transformer Featherless Exclusive Cold

FRPO/qwen3-4b-a1_base-k1-cNone-clip0.2-mb4-eta100-bs256x5-n4 is a 4 billion parameter language model derived from the Qwen3-4B architecture. Developed by FRPO as part of the KL-in-LLM-RL experiments, this model is an RL fine-tuned checkpoint. It is specifically designed for research and development in reinforcement learning applications, utilizing the verl training framework.

Loading preview...

Model Overview

This model, FRPO/qwen3-4b-a1_base-k1-cNone-clip0.2-mb4-eta100-bs256x5-n4, is a 4 billion parameter language model based on the Qwen/Qwen3-4B architecture. It represents a specific checkpoint (global_step_200) from the KL-in-LLM-RL / FRPO experimental series, which focuses on applying Reinforcement Learning (RL) techniques to large language models.

Key Characteristics

  • Base Model: Qwen3-4B, a robust foundation for language understanding and generation.
  • Fine-tuning Method: Utilizes Reinforcement Learning (RL) with the verl framework, specifically employing the FRPO (Fictitious Reinforcement Policy Optimization) approach.
  • Parameter Count: 4 billion parameters, offering a balance between performance and computational efficiency.
  • Weights: Provided in fp32 safetensors format, directly as saved during the training process without further post-processing.
  • Configuration: The specific run configuration for this checkpoint is encoded within its repository name, indicating parameters like clip, mb, eta, bs, and n.

Intended Use Cases

This model is primarily suited for:

  • RL Research: Ideal for researchers exploring the application of Reinforcement Learning to language models, particularly within the KL-in-LLM-RL paradigm.
  • Experimental Development: Useful for developers and researchers who need an RL-fine-tuned Qwen3-4B checkpoint for comparative studies or building upon the FRPO methodology.
  • Understanding RL Impact: Provides a concrete example of an LLM fine-tuned with specific RL algorithms, allowing for analysis of their effects on model behavior and capabilities.

Finetunes
1 models