trinityomnis/OpenThinkerAgent-32B-SFT-316

TEXT GENERATIONPricing:Input $0.408 / Cached $0.0816 / Output $1.972Concurrent Unit Cost:2Model Size:32BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 7, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

OpenThinkerAgent-32B-SFT-316 is a 32 billion parameter language model developed by OpenThoughts, fine-tuned from Qwen3-32B. It is specifically optimized for agentic tasks, leveraging a supervised fine-tuning (SFT) approach on a curated dataset of agent trajectories. This model excels in complex problem-solving within environments like SWE-Bench and Terminal-Bench, demonstrating enhanced performance in agent-based reasoning and execution compared to its base model.

Loading preview...

OpenThinkerAgent-32B-SFT-316 Overview

OpenThinkerAgent-32B-SFT-316 is a 32 billion parameter model developed by OpenThoughts, derived from the Qwen3-32B architecture. It has undergone supervised fine-tuning (SFT) using the OpenThoughts-Agent-SFT-316 dataset, which comprises 316 high-quality agent trajectories. This specialized training focuses on enhancing the model's capabilities in agentic problem-solving.

Key Capabilities

  • Agentic Task Performance: Demonstrates improved performance on agent-specific benchmarks such as OpenThoughts-TBLite and Terminal-Bench 2.0, surpassing its base model, Qwen3-32B.
  • Complex Problem Solving: Trained on diverse task sources including SWE-Smith, StackExchange, and IssueTasks, enabling it to navigate and resolve intricate problems.
  • Supervised Fine-Tuning: Utilizes a full-parameter SFT approach with a GLM-4.7-AWQ teacher model, ensuring high-quality trajectory learning.
  • Extensive Context Window: Supports a context length of 32768 tokens, facilitating the processing of long and complex agentic interactions.

Good For

  • Developing AI Agents: Ideal for applications requiring models to act as autonomous agents, performing multi-step tasks and interacting with environments.
  • Research in Agentic AI: Provides a strong baseline for researchers exploring agentic capabilities, dataset curation, and fine-tuning methodologies.
  • Automated Problem Solving: Suitable for tasks involving code generation, debugging, and general problem-solving within terminal or software engineering environments.

Finetunes
1 models