FRPO/qwen3-1.7b-a3_onpolicy-k1-cNone-clip0.2-mb1-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-a3_onpolicy-k1-cNone-clip0.2-mb1-eta100-bs256x5-n2 is a 2 billion parameter language model based on the Qwen3-1.7B architecture, developed by FRPO. This model has been fine-tuned using Reinforcement Learning (RL) with the FRPO algorithm, specifically for on-policy optimization. It features a 32768-token context length and is designed for tasks benefiting from RL-based performance enhancements.

Loading preview...

Overview

This model, FRPO/qwen3-1.7b-a3_onpolicy-k1-cNone-clip0.2-mb1-eta100-bs256x5-n2, is a 2 billion parameter language model derived from the Qwen3-1.7B base architecture. It has undergone Reinforcement Learning (RL) fine-tuning using the FRPO algorithm, specifically configured for on-policy learning experiments.

Key Characteristics

  • Base Model: Qwen/Qwen3-1.7B, providing a robust foundation.
  • Fine-tuning Method: Utilizes the FRPO algorithm for RL-based optimization, trained with the verl framework.
  • Parameter Count: Approximately 2 billion parameters.
  • Context Length: Supports a substantial context window of 32768 tokens.
  • Weights: Provided in fp32 safetensors format, directly from the trainer without post-processing.
  • Configuration: The specific run configuration for its RL training is encoded within the repository name itself.

When to Consider This Model

This model is particularly suited for research and applications where:

  • You are exploring the effects of Reinforcement Learning from Human Feedback (RLHF) or similar RL-based fine-tuning approaches.
  • Your use case benefits from a model optimized through on-policy RL algorithms.
  • You require a model with a large context window (32768 tokens) for processing extensive inputs.
  • You are working with the Qwen3-1.7B family and need an RL-tuned variant.