spade-rl/SPADE-Qwen3-8B-Games

TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 12, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

SPADE-Qwen3-8B-Games is an 8 billion parameter language model developed by spade-rl, fine-tuned from Qwen3-8B. It is specifically designed for game environments, functioning as both an Environment Designer that creates executable game environments and a Reasoning Agent that solves them. This model utilizes a unique curriculum learning approach where the Designer generates environments at the frontier of the Agent's current capabilities, ensuring adaptive and progressive training. With a context length of 32,768 tokens, it excels in dynamic game-related tasks.

Loading preview...

SPADE-Qwen3-8B-Games Overview

SPADE-Qwen3-8B-Games is an 8 billion parameter language model, fine-tuned by spade-rl from the base Qwen3-8B architecture. This model is specialized for game environments, operating in a dual-role capacity as both an Environment Designer and a Reasoning Agent.

Key Capabilities

  • Adaptive Environment Generation: The model's Designer component is rewarded for creating executable environments that push the boundaries of what the Agent can currently solve, fostering a dynamic and self-improving curriculum.
  • Problem Solving in Games: The Reasoning Agent component is trained to effectively solve the environments generated by the Designer.
  • Extended Context Window: Supports a substantial context length of 32,768 tokens, enabling complex interactions and understanding within game scenarios.
  • Curriculum Learning: Implements a unique training methodology where the learning curriculum adapts to the Agent's evolving policy, rather than being fixed.

Good For

  • Game Development and AI: Ideal for researchers and developers working on AI agents that need to interact with, design, or solve problems within game environments.
  • Dynamic Task Generation: Suitable for applications requiring an AI to not only solve problems but also to generate challenging, solvable tasks.
  • Reinforcement Learning Research: Provides a robust platform for exploring adaptive curriculum generation and multi-agent roles in reinforcement learning settings.

For more detailed information, refer to the original paper describing the SPADE framework.

Finetunes
1 models