laion/a3-rl-DCAgent_selfinstruct-naive-sandboxes-2-verified-70-8B

TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 4, 2026Architecture:Transformer Featherless Exclusive Cold

The laion/a3-rl-DCAgent_selfinstruct-naive-sandboxes-2-verified-70-8B is an 8 billion parameter language model developed by laion, fine-tuned using Reinforcement Learning (RL) with the SkyRL GRPO algorithm. It is based on the GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink model and trained on the DCAgent/selfinstruct-naive-sandboxes-2-verified dataset. This model is specifically optimized for tasks related to self-instruction and agent-based environments, leveraging its RL training for improved performance in sandbox scenarios.

Loading preview...

Model Overview

This model, a3-rl-DCAgent_selfinstruct-naive-sandboxes-2-verified-70-8B, is an 8 billion parameter language model developed by laion. It is a fine-tuned version of the laion/GLM-4_7-swesmith-sandboxes-with_tests-oracle_verified_120s-maxeps-131k-fixthink base model, specifically enhanced through Reinforcement Learning (RL).

Training Details

The model underwent RL training using the SkyRL GRPO algorithm. The training utilized a substantial setup of 56 GPUs across 14 nodes, with a fan-out K=4. The RL dataset employed for fine-tuning was DCAgent/selfinstruct-naive-sandboxes-2-verified. Training traces, including the last episode of each trial, are available as a companion dataset: open-athena/a3-rl-DCAgent_selfinstruct-naive-sandboxes-2-verified.

Key Characteristics

  • RL-tuned: Enhanced through Reinforcement Learning, making it suitable for agent-based tasks.
  • Base Model: Built upon the GLM-4_7-swesmith series, indicating a foundation in robust language understanding.
  • Dataset Specificity: Trained on a dataset focused on self-instruction and sandbox verification, suggesting capabilities in structured problem-solving within defined environments.

Potential Use Cases

This model is particularly well-suited for applications requiring:

  • Agent-based simulations: Where an AI agent needs to learn and adapt within a sandbox or simulated environment.
  • Self-instruction scenarios: Tasks that involve generating instructions or learning from self-generated data.
  • Verified task execution: Environments where actions need to be verified against specific conditions or rules.

Finetunes
1 models