AI Daily · Sep 17, 2026: GPT-6 Astra Pro free for US clinicians; AlphaGenome Atlas opens for disease genetics
OpenAI offers free GPT-6 Astra Pro to verified US clinicians; Google DeepMind opens AlphaGenome Atlas for disease genetics. StepFun also releases stepaudio3 music generation model.
Pick a date
-
Free GPT-6 Astra Pro Access for U.S. Clinicians
OpenAI's 'ChatGPT for Clinicians' offers verified U.S. clinicians free GPT-6 Astra Pro access, with signup available immediately.
-
Google DeepMind Opens AlphaGenome Atlas for Disease Genetics
Google DeepMind made AlphaGenome Atlas freely accessible, a database designed to help researchers decode the genetic causes of disease, already being used by researchers at Broad Institute and universities. The atlas maps over 2500 regulatory patterns, identifying DNA sequences acting as switches and volume dials for gene activity.
-
StepFun launches StepAudio 3 Music foundation model for full song generation
StepFun and ACE Studio introduced StepAudio 3 Music, converting text prompts and lyrics into complete songs. Users can specify genre, mood, vocal character, instruments, key, BPM, structure, and arrangement. A Hugging Face Space, API docs, and thesis are available.
-
Union Alpha free in Cline with 256k context and multimodal support
Cline announced that Union Alpha, a stealth AI model with 256k context and multimodal capabilities, is now available for free, supporting tool calling. Users can install globally via npm and select it in the Free tier; free for the next week with recent speed improvements.
-
OpenRouter launches openrouter:shell for hosted code execution
OpenRouter introduced openrouter:shell, letting any model run code in a hosted Linux container by adding one tool to a Responses API request. Since soft launch, sandboxes have been launched over 18,000 times.
-
OpenAI publishes framework for disclosing model misalignment incidents
OpenAI published a new framework for tracking, investigating, and publicly disclosing model misalignment cases, including criteria and timelines for disclosure even when behavior is not fully explained or mitigated.
-
Anthropic merges Claude Cowork and chat into unified experience with docs editing
Anthropic announced merging Claude Cowork and chat into a unified experience, allowing handoff and continued work after closing the laptop. Claude now works directly inside documents and slide decks. Design, Slides, and Docs are in beta for paid plans.
-
Grok Voice Speech Model Now Live on fal
Grok Voice is now available on fal, responding in 0.70 seconds with word-level timestamps, TTS in 30+ voices across 25+ languages, and tool calls completed before the sentence ends.
-
Qwen-Image 2.1 Open Sourcing with 50 Early Access Spots
Qwen Developers announced Qwen-Image 2.1 will be open sourced, opening 50 early access spots for experienced creators and developers; selected participants must publish at least one original showcase.