Published AI-Buzz coverage
Catch up on what changed.
27 selected developments, newest announcements first. Open a story for the explanation, supporting facts and related tools.
Which of these tools can you access?
One row per covered development, using the same access statement as its story. Different products, plans and previews are separate scopes; they are not a ranking. These are provider descriptions checked on the stated dates, not a history of tested availability. Later availability is unknown until reviewed.
| Tool and development | Access and important limits | Source and dates |
|---|---|---|
| Muse Meta | Meta says US access is rolling out on iOS, Android and the web. Reliability has not been tested by AI-Buzz. | Meta’s Muse announcement Announcement Published |
| ChatGPT Images 2.5 OpenAI | OpenAI describes rollout to all tiers across ChatGPT, ChatGPT Work and Codex on web, desktop and mobile. Flare and Sunburst are separate API options. Account access and output quality have not been tested here. | OpenAI’s Images 2.5 announcement Announcement Published |
| Lyria 3.5 in Gemini | Google says the update is available to Gemini users globally on web and mobile, and names Flow Music, AI Studio, Vids and the API. The announcement does not specify quotas or a universal maximum track length. | Google’s Lyria 3.5 announcement Announcement Published |
| GitHub Copilot CLI · Project HydraFusion Microsoft | Research preview on all Copilot plans through /experimental, subject to organization CLI policy. Usage includes tokens from every model invoked at its standard rate. GitHub recommends first-turn, single-prompt tasks for this preview. | GitHub’s HydraFusion research preview Announcement Supporting sources: Published |
| Gmail Live, Docs Live and Keep Live | The article says Gmail and Keep are rolling out to Google AI Plus, Pro and Ultra; Docs to Pro and Ultra. Workspace business access is “coming soon.” | Google’s voice-feature announcement Announcement Supporting sources: Published |
| Funes Hugging Face | Developer tool with local search, embedding and reranking. Local memory does not make the coding assistant’s reasoning offline. Optional dataset synchronization uploads session material; existing repositories retain their visibility. | Hugging Face’s Funes announcement Announcement Supporting sources: Published |
| MAI-Transcribe-2 Microsoft | Azure Speech public preview requires an Azure subscription and Speech resource. Its documentation gives no SLA and does not recommend production workloads. Regional availability is limited; transcript accuracy has not been tested here. | Microsoft’s MAI-Transcribe-2 announcement Announcement Supporting sources: Published |
| GPT-6 Astra OpenAI | OpenAI’s initial rollout was limited. Its later announcement names ChatGPT Work, Codex and API access; Enterprise administrators must enable it. AI-Buzz has not tested an account or completed tasks with it. | OpenAI’s Astra launch Announcement Supporting sources: Published |
| GWM Worlds 2 Runway | Research preview. The announcement offers a contact form; it does not establish public access. | Runway’s research and demos Announcement Published |
| Claude Fable 5.1 and Mythos 5.1 Anthropic | Anthropic says Fable is available across its platforms, AWS, Google Cloud and Azure. Mythos is limited to selected vetted US organizations; wider access is planned. Enterprise Frontier Safeguards are a later rollout. | Anthropic’s Fable and Mythos announcement Announcement Supporting sources: Published |
| @huggingface/kernels for WebGPU Hugging Face | Developer library in preview under Apache 2.0, not a consumer chat app. It needs compatible WebGPU support, which varies with browser, operating system, GPU and drivers. | Hugging Face’s WebGPU kernels release Announcement Published |
| ChatGPT for Healthcare OpenAI | Workspace administrators must enable the EHR integration; eligible Enterprise customers go through their account team. Eligible US ChatGPT for Clinicians accounts can use the public-data plugin, but individual accounts do not get the EHR integration. | OpenAI’s healthcare connections announcement Announcement Published |
| Computer · Hybrid Compute on Mac Perplexity | Pro, Max or Enterprise required, plus an Apple silicon Mac with macOS 15 or later and at least 24 GB of unified memory. These feature requirements are stricter than a general Mac-app download requirement. | Perplexity’s Mac hybrid-compute announcement Announcement Supporting sources: Published |
| Solaris Runway | Research demonstration with an early-access request form. Stable text, factual reliability, long sessions and accessibility interfaces remain limitations. This does not establish general public access or dependable application software. | Runway’s Solaris research announcement Announcement Published |
| Model Hardware Standard Anthropic | Limited research preview for scientific labs and advanced manufacturers; application required. Open sourcing is planned. The announcement does not establish dependable autonomous laboratory operation. | Anthropic’s hardware-standard preview Announcement Published |
| Claude research-lab offer Anthropic | A principal investigator or equivalent must verify the lab. Standard seats are free; premium seats with five times the usage cost US$15 per month. Biology and chemistry researchers remain limited to Opus-class models; Fable blocks professional biology and drug-development queries. Places are finite and application is not acceptance. | Anthropic’s research-access offer Announcement Published |
| Gemini Omni 1.1 Flash | Google describes developer rollout through AI Studio and the Agent Platform API, Flow access for AI Plus, Pro and Ultra, and scene extension in Gemini. Flow is 18+; features vary by tier, platform and region. | Google’s Omni 1.1 Flash announcement Announcement Supporting sources: Published |
| Gemini Live with Spark and Daily Brief | Spark requires Google AI Pro or higher; Daily Brief requires AI Plus or higher. Users choose connected apps in Personal Intelligence settings. The announcement does not establish every country or account’s access. | Google’s Gemini Live announcement Announcement Published |
| Portable Computer on DGX Spark Perplexity | Pro or Max and DGX Spark hardware are required. Initial support is Linux on DGX Spark; Windows and broader RTX support are future work. Local execution uses no Computer credits, but the subscription and hardware are not free. | NVIDIA’s dated Portable Computer entry Announcement Supporting sources: Published |
| Computer licensed-data connectors Perplexity | Pro, Max or Enterprise is required, plus an eligible subscription to each relevant data provider. The announcement does not make those databases free or remove their access terms. | Perplexity’s licensed-data announcement Announcement Supporting sources: Published |
| ElevenLabs CLI v1 ElevenLabs | Developer tool requiring OAuth or API-key authentication. The August 24 article was updated September 6, so its current body does not establish every feature present at the original launch. | ElevenLabs’ CLI v1 announcement Announcement Supporting sources: Published |
| Computer in Email Perplexity | Available to Computer users with a verified sender address; this does not mean every free Perplexity account. Replies go to that sender. Reply-all collaboration is described as forthcoming. | Perplexity’s August 24 email update Announcement Supporting sources: Published |
| ElevenLabs hosted MCP ElevenLabs | OAuth connects an ElevenLabs account and grants workspace capabilities, including agent changes. Regional workspaces have different accounts and URLs. The documented connection does not expose the entire audio API. | ElevenLabs’ August 22 retirement notice Announcement Supporting sources: Published |
| Agentic Search · Search Toolkit Mistral | Developer tooling requiring document setup; the Python toolkit lists Python 3.12 or later. Mistral describes cloud and on-premises options. Accuracy improvements are vendor claims, not measurements by AI-Buzz. | Mistral’s Agentic Search announcement Announcement Supporting sources: Published |
| ChatGPT for Teens OpenAI | OpenAI says declared ages 13–17, and accounts estimated to belong to someone under 18, receive the teen experience. The announcement does not provide a country-by-country rollout or plan matrix. Safeguards are not guarantees of safe outcomes. | OpenAI’s teen-experience announcement Announcement Supporting sources: Published |
| Claude text watermark and detector Anthropic | The detection API is a private preview for eligible organizations, with an interest form. Anthropic describes future-model watermarking and a later rollout for older models; this does not mean every Claude response is already marked. | Anthropic’s text-watermark explanation Announcement Published |
| Dubbing v2 API ElevenLabs | An API key is required. Each project has a minimum charge for one language; further target languages cost extra. Transcript editing and regeneration endpoints are Enterprise-only, and regeneration is charged like generation. | ElevenLabs’ August 10 API changelog Announcement Supporting sources: Published |
Publication and review days use America/New_York. Source dates use the source’s own date. How AI-Buzz handles sources, date uncertainty and corrections.
September 2026
Meta puts an agent inside a conversation.
Muse
Access Meta says US access is rolling out on iOS, Android and the web. Reliability has not been tested by AI-Buzz.
Images 2.5 lets you sketch the change you want.
ChatGPT Images 2.5
Access OpenAI describes rollout to all tiers across ChatGPT, ChatGPT Work and Codex on web, desktop and mobile. Flare and Sunburst are separate API options. Account access and output quality have not been tested here.
Lyria 3.5 gives music prompts more direction.
Lyria 3.5 in Gemini
Access Google says the update is available to Gemini users globally on web and mobile, and names Flow Music, AI Studio, Vids and the API. The announcement does not specify quotas or a universal maximum track length.
Copilot CLI can coordinate a task across several models.
GitHub Copilot CLI · Project HydraFusion
Access Research preview on all Copilot plans through /experimental, subject to organization CLI policy. Usage includes tokens from every model invoked at its standard rate. GitHub recommends first-turn, single-prompt tasks for this preview.
Google adds conversation to Gmail, Docs and Keep.
Gmail Live, Docs Live and Keep Live
Access The article says Gmail and Keep are rolling out to Google AI Plus, Pro and Ultra; Docs to Pro and Ultra. Workspace business access is “coming soon.”
Funes makes earlier coding-agent sessions searchable.
Funes
Access Developer tool with local search, embedding and reranking. Local memory does not make the coding assistant’s reasoning offline. Optional dataset synchronization uploads session material; existing repositories retain their visibility.
MAI-Transcribe-2 adds speaker labels and word timings.
MAI-Transcribe-2
Access Azure Speech public preview requires an Azure subscription and Speech resource. Its documentation gives no SLA and does not recommend production workloads. Regional availability is limited; transcript accuracy has not been tested here.
Astra takes on tasks across existing applications.
GPT-6 Astra
Access OpenAI’s initial rollout was limited. Its later announcement names ChatGPT Work, Codex and API access; Enterprise administrators must enable it. AI-Buzz has not tested an account or completed tasks with it.
Runway shows generated worlds you can steer.
GWM Worlds 2
Access Research preview. The announcement offers a contact form; it does not establish public access.
Fable 5.1 and Mythos 5.1 come with different access rules.
Claude Fable 5.1 and Mythos 5.1
Access Anthropic says Fable is available across its platforms, AWS, Google Cloud and Azure. Mythos is limited to selected vetted US organizations; wider access is planned. Enterprise Frontier Safeguards are a later rollout.
Browser AI gets reusable GPU building blocks.
@huggingface/kernels for WebGPU
Access Developer library in preview under Apache 2.0, not a consumer chat app. It needs compatible WebGPU support, which varies with browser, operating system, GPU and drivers.
Healthcare workspaces can connect patient records.
ChatGPT for Healthcare
Access Workspace administrators must enable the EHR integration; eligible Enterprise customers go through their account team. Eligible US ChatGPT for Clinicians accounts can use the public-data plugin, but individual accounts do not get the EHR integration.
Perplexity Computer adds local models on supported Macs.
Computer · Hybrid Compute on Mac
Access Pro, Max or Enterprise required, plus an Apple silicon Mac with macOS 15 or later and at least 24 GB of unified memory. These feature requirements are stricter than a general Mac-app download requirement.
August 2026
Solaris generates an interface as you use it.
Solaris
Access Research demonstration with an early-access request form. Stable text, factual reliability, long sessions and accessibility interfaces remain limitations. This does not establish general public access or dependable application software.
A hardware standard moves into supervised laboratory trials.
Model Hardware Standard
Access Limited research preview for scientific labs and advanced manufacturers; application required. Open sourcing is planned. The announcement does not establish dependable autonomous laboratory operation.
Verified research labs can apply for subsidized Claude access.
Claude research-lab offer
Access A principal investigator or equivalent must verify the lab. Standard seats are free; premium seats with five times the usage cost US$15 per month. Biology and chemistry researchers remain limited to Opus-class models; Fable blocks professional biology and drug-development queries. Places are finite and application is not acceptance.
Omni 1.1 Flash adds more control over a generated video.
Gemini Omni 1.1 Flash
Access Google describes developer rollout through AI Studio and the Agent Platform API, Flow access for AI Plus, Pro and Ultra, and scene extension in Gemini. Flow is 18+; features vary by tier, platform and region.
Gemini Live can hand spoken requests to background tasks.
Gemini Live with Spark and Daily Brief
Access Spark requires Google AI Pro or higher; Daily Brief requires AI Plus or higher. Users choose connected apps in Personal Intelligence settings. The announcement does not establish every country or account’s access.
Portable Computer runs agent work on a DGX Spark.
Portable Computer on DGX Spark
Access Pro or Max and DGX Spark hardware are required. Initial support is Linux on DGX Spark; Windows and broader RTX support are future work. Local execution uses no Computer credits, but the subscription and hardware are not free.
Computer connects to more licensed business databases.
Computer licensed-data connectors
Access Pro, Max or Enterprise is required, plus an eligible subscription to each relevant data provider. The announcement does not make those databases free or remove their access terms.
Voice-agent configuration moves into the terminal.
ElevenLabs CLI v1
Access Developer tool requiring OAuth or API-key authentication. The August 24 article was updated September 6, so its current body does not establish every feature present at the original launch.
Computer tasks can start from an email thread.
Computer in Email
Access Available to Computer users with a verified sender address; this does not mean every free Perplexity account. Replies go to that sender. Reply-all collaboration is described as forthcoming.
ElevenLabs retires its local MCP tools.
ElevenLabs hosted MCP
Access OAuth connects an ElevenLabs account and grants workspace capabilities, including agent changes. Regional workspaces have different accounts and URLs. The documented connection does not expose the entire audio API.
Mistral’s document search can investigate its first results.
Agentic Search · Search Toolkit
Access Developer tooling requiring document setup; the Python toolkit lists Python 3.12 or later. Mistral describes cloud and on-premises options. Accuracy improvements are vendor claims, not measurements by AI-Buzz.
ChatGPT adds study defaults and family controls for teens.
ChatGPT for Teens
Access OpenAI says declared ages 13–17, and accounts estimated to belong to someone under 18, receive the teen experience. The announcement does not provide a country-by-country rollout or plan matrix. Safeguards are not guarantees of safe outcomes.
Claude’s watermark detector enters a restricted preview.
Claude text watermark and detector
Access The detection API is a private preview for eligible organizations, with an interest form. Anthropic describes future-model watermarking and a later rollout for older models; this does not mean every Claude response is already marked.
Dubbing v2 reaches the API.
Dubbing v2 API
Access An API key is required. Each project has a minimum charge for one language; further target languages cost extra. Transcript editing and regeneration endpoints are Enterprise-only, and regeneration is charged like generation.