Tech • AI • Robotics • Game
AI News: New ChatGPT Limits, iMessage Dot, Gemini 4 Carbon & A Terrifying Voice Model
TL;DR
OpenAI, Google, Anthropic and xAI unveiled a wave of product and model updates, led by higher ChatGPT usage limits, texting access for Dots, Google’s benchmark-leading Gemini 4 Carbon, and new tools for voice, automation and document creation.
KEY POINTS
OpenAI raises or preserves limits
OpenAI reversed a planned cut to the $200 plan and instead expanded allowances for some higher tiers. Users on the $100 and $500 pro plans now get double their previous Work and Codex usage, while the $200 plan keeps its current limits rather than dropping. Plus subscribers are not included in the increase.
Dots moves into text messaging
Dots, OpenAI’s proactive assistant, can now be contacted directly from a phone’s messaging app after number verification. The feature is in beta for U.S. pro users and supports image uploads, location sharing and voice messages. That allows tasks such as asking for nearby recommendations, sending photos for analysis, or dictating requests while driving or from a cellular smartwatch.
New OpenAI model targets computer use
OpenAI released GPT 6.1 Terra, positioned between faster low-end models and more powerful premium ones. The model is designed for screen understanding, app actions and computer-use tasks including basic video editing. It performed well on straightforward jobs such as trimming silences, but more advanced edits and interface decisions remained less reliable than stronger models such as 6.1 Soul.
Google introduces Gemini 4 Carbon
Google announced Gemini 4 Carbon, a new frontier model that it says tops major benchmarks against GPT 6 Astra, Claude Opus 5.5 and Claude Fable. The model is not yet broadly available and is being limited at first while security issues are tested and patched. A wider release could follow the pattern of earlier launches, which reached the public within about two weeks.
Block reveals highly realistic voice model
Block, the company behind Square and Cash App, previewed a research voice model called Desa. Unlike conventional assistants that convert speech to text and back again, Desa was trained directly on human conversation audio. The result is unusually natural speech, with pauses, interruptions and emotional cadence that make demos sound strikingly human.
Claude gets app upgrades and IPO details emerge
Anthropic refreshed Claude on mobile and desktop, adding better layouts for larger devices and support for multiple signed-in accounts on iPhone and Android. Voice mode is now available broadly in Apple CarPlay. The company also filed to go public on the NASDAQ under ticker ANT, reporting $16.3 billion in first-half 2026 revenue and an $80.7 billion net loss, though about $72 billion was described as a non-cash accounting item.
Claude Duet expands in the U.S.
Anthropic said Claude Duet, its autonomous assistant product, is being opened first to U.S. Max users and then to U.S. Pro users. No launch date was given for other countries. The staged rollout mirrors the broader industry pattern of testing higher-autonomy features on paid tiers before wider release.
Grokbot adds calls, video and password vault
Grokbot is expected to gain the ability to call businesses or individuals and then return a summary of the conversation. It has also added direct video generation and a secure vault for saved credentials, contact details and payment information. The vault is designed so the bot can autofill logins without directly viewing stored passwords.
Google expands interactive worlds and document editing
Google also showcased Genie 4, an interactive world model that generates explorable environments in real time with stronger visuals and audio than Genie 3. Separately, Gemini can now generate and edit PDFs, presentations and spreadsheets with live previews, including converting handwritten notes into formatted study materials. Another update lets users share their Gemini avatars with friends and family for collaborative video generation.
CONCLUSION
The week’s releases point to a more practical phase of AI competition, with companies focusing less on novelty alone and more on higher limits, easier access and tools that fit directly into daily workflows. The biggest near-term impact is likely to come from messaging-based assistants, improved automation and broader access to stronger models.
Ask a question