Models are considered released in open-source if the weights are available to some members of the public beyond a closed beta, irrespective of license.
The AI released must have a language output modality, ie it must be a language model. A model that only generates images or videos, or a robotic model, would not count.
See also:
/Bayesian/when-will-google-release-gemini-3
/Bayesian/when-will-xai-release-grok-5
/Bayesian/when-will-openai-release-gpt6
/Bayesian/when-will-anthropic-release-claude-qZd5QEQ225
/Bayesian/when-will-deepseek-release-r2
/Bayesian/when-will-deepseek-release-v4
/Bayesian/when-will-meta-release-llama-5-6h9UypqOdp
/Bayesian/when-will-alibaba-release-qwen-4
/Bayesian/when-will-openai-release-a-browser
/Bayesian/when-will-xai-release-grokipedia
/Bayesian/when-will-openai-release-their-next (this market)
1,000@Bayesian i'll bet when u vote https://manifold.markets/jim/straight-mens-would-you-date-a-tran?r=amlt