What changed, who could use it, and the original sources.
In his Day 2.1 post, Tibo said Auto-review is free for all users signed in through a ChatGPT account. It can be enabled under Settings → Permissions → Auto-review, reducing repetitive approval decisions.
Access at announcementOnly the readable portion of the original post was verified. This update does not establish quota accounting, every high-risk approval rule, or unlimited general Codex usage.
Cursor added remote control of local agents through its iOS app. You can inspect a running agent and send it messages after pairing the computer; the agent continues executing on that computer.
Access at announcementRequires the Cursor iOS app and a paired, available desktop computer. To use the keep-awake setting, the computer must be plugged in with its lid open. No new subscription price is established by this announcement.
The stable Diffusers 0.41.0 release adds Qwen Image 2.1 text-to-image and editing integrations, including multi-reference editing, RGBA output and LoRA support. This is a library release, not evidence of a new Qwen model announcement.
Access at announcementThe official GitHub release was published October 6. Model-specific weights, licenses and compute requirements remain separate from the library version.
Google DeepMind released EmbeddingGemma 2, an open embedding model for text and code, images, video, and audio. Its modular architecture has about 740M parameters and supports a shared 8,192-token input budget.
Access at announcementPublic weights are available on Hugging Face and Kaggle under Apache 2.0; the model card also requires adherence to the Gemma Prohibited Use Policy. Model Garden availability is forthcoming. Media modalities share the same input budget.
Mistral officially announced Large 4 and opened an API preview. The current model documentation lists a 1M-token context. Weights are planned for late October, so the preview is not evidence that public downloads are already available.
Access at announcementAvailable through the API as a public preview. Check current model documentation for access and billing; planned open weights are kept separate from current availability.
Aleph Alpha announced Kolibri 1 on October 5; the model card dates availability to October 3. This MoE reasoning model provides Apache 2.0 weights, with 78.1B total and 3.46B active parameters, explicit reasoning mode and tool calling.
Access at announcementWeights are available under Apache 2.0. Local deployment requires suitable compute; announcement and initial availability dates differ.
