lausannequants/OpenThinker-7B

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 2, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

OpenThinker-7B is a 7.6 billion parameter instruction-tuned causal language model developed by lausannequants, fine-tuned from Qwen2.5-7B-Instruct. It is specifically optimized for reasoning tasks, leveraging the OpenThoughts-114k dataset which distills DeepSeek-R1. This model demonstrates improved performance on various reasoning benchmarks, including AIME24, MATH500, and GPQA-Diamond, making it suitable for complex analytical applications.

Loading preview...

OpenThinker-7B: A Reasoning-Optimized Language Model

OpenThinker-7B is a 7.6 billion parameter language model developed by lausannequants, built upon the Qwen2.5-7B-Instruct architecture. Its primary distinction lies in its fine-tuning on the OpenThoughts-114k dataset, a high-quality dataset derived from distilling DeepSeek-R1. This specialized training aims to enhance the model's reasoning capabilities across various domains.

Key Capabilities & Performance

OpenThinker-7B shows notable improvements in reasoning benchmarks compared to its predecessor, Bespoke-Stratos-7B. Evaluated using the open-source Evalchemy tool, it achieves:

  • AIME24: 31.3 (vs. 22.7 for Bespoke-Stratos-7B)
  • MATH500: 83.0 (vs. 79.6 for Bespoke-Stratos-7B)
  • GPQA-Diamond: 42.4 (vs. 38.9 for Bespoke-Stratos-7B)
  • LCBv2 Hard: 6.5 (vs. 0.8 for Bespoke-Stratos-7B)

These metrics highlight its enhanced proficiency in complex problem-solving and analytical tasks. The model was trained for 20 hours on four 8xH100 nodes, emphasizing a robust training procedure.

Open-Source Commitment

Lausannequants maintains a strong commitment to open science, providing:

  • Open Weights: The model weights are publicly available.
  • Open Data: The OpenThoughts-114k dataset is accessible.
  • Open Code: Data generation, evaluation (Evalchemy), and training code are all open-source on GitHub.

Intended Use Cases

OpenThinker-7B is particularly well-suited for applications requiring strong reasoning and analytical skills, such as:

  • Mathematical problem-solving
  • Complex question answering
  • Logical inference tasks
  • Research and academic applications where robust reasoning is critical.

Finetunes
1 models