By September 30, 2025, will an open-source AI model be released that surpasses OpenAI's o1 preview (or full) model on major performance benchmarks, with its weights available for commercial use under an attribution license similar to Meta’s Llama?
Time Frame: The model must be released by the end of the second quarter of 2025 (September 30, 2025).
Open-Source Model Criteria:
Weights Availability: The model’s weights must be publicly accessible and available for commercial use, allowing businesses and developers to freely integrate it into their applications.
Attribution and Licensing: The licensing terms must require attribution, in line with prior releases such as Meta’s Llama, meaning users must credit the creators but can use the model commercially.
Performance Criteria: The model must outperform OpenAI’s o1 preview (or full) model on at least two widely recognized AI benchmarks or https://lmarena.ai/
(inspired from https://manifold.markets/JohnL/by-the-end-of-q1-2025-will-an-open)
1,000🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ31 | |
| 2 | Ṁ9 | |
| 3 | Ṁ8 | |
| 4 | Ṁ8 | |
| 5 | Ṁ8 |
RESOLUTION (see https://manifold.markets/JohnL/by-the-end-of-q1-2025-will-an-open) Qwen's QwQ has met all criteria for the challenge.
- It beats 01-preview on both AIME and MATH-500
- Is available for commercial use under the Apache 2.0 license
- Has weights available on Hugging Face
https://qwenlm.github.io/blog/qwq-32b-preview/