ChatGPT Voice gets GPT-6, plugin support, and document creation

Talking to an AI on your phone is nothing new, but having it build an entire slide deck or crunch numbers into a spreadsheet just from your voice is a whole different ballgame.

OpenAI announced a major global rollout for ChatGPT Voice. This is not just another conversational tweak; it is a full-fledged push to turn voice into an actionable workspace tool.

Building docs and decks by voice

The headline feature here is deep integration with external tools and productivity setups. ChatGPT Voice now connects directly with plugins including email, calendar, and Slack.

Even better, it now works inside ChatGPT Work across both web and mobile. Users can draft documents, design decks, set up simple sites, create spreadsheets, and tackle multi-step browser tasks purely by speaking commands.

According to coverage from Forward Future, Plus and Pro subscribers can now jump into the Work tab on mobile to dictate document drafts or summarize messy Slack threads on the fly, keeping their hands completely free.

Powered by GPT-6 Astra, Sol, and Luna

Backing these new capabilities is OpenAI's fresh GPT-6 model tier, split into three specific categories: Astra, Sol, and Luna.

Astra sits at the top as the flagship engine with the highest reasoning benchmark scores, but it also carries the steepest price tag. According to technical breakdowns from reviewers, the mid-tier Sol scores only slightly below Astra and top-end Claude models while coming in over three times cheaper.

Then there is Luna, the ultra-lightweight option built for pure speed and cost efficiency. Reviewers note that running a task on Luna costs around 7 cents, making it more than 100 times cheaper than premier Claude models. Paid subscribers can access both Sol and Luna in work and coding environments, while free users can already test Luna on desktop.

Why it matters

We think this update represents a smart shift from treating voice as a novelty chat companion to deploying it as a practical remote control for AI agents.

For desk workers and mobile professionals, the value is obvious: being able to turn raw thoughts into structured presentations or clear email updates during a commute is an immediate time-saver. By pairing lightweight models like Luna for fast inputs with heavier engines for complex generation, OpenAI is keeping infrastructure costs realistic while expanding scale.

Across the industry, this intensifies pressure on rivals like Anthropic, which has also been working to streamline mobile-to-desktop handoffs. However, handling intricate spreadsheets and precision slide formatting purely through voice still has obvious friction. If the model misinterprets an instruction, refining the output by voice alone might still test a user's patience compared to grabbing a keyboard.