ICYMI 👀: OpenAI Platform now has a "Decisions" API playground.

Users can test OpenAI's decision model with the "Predicate," "Choice," and "Score" options, which were released recently.

Probability True 🤖

DAILY AI BRIEF 🗞 — Oct 20

OPENAI 🔥:
- ChatGPT Slides now supports real-time collaborative editing, comments and on-slide edits for Pro, Business and Enterprise plans.
- OpenAI Platform added a "Decisions" API playground for testing its decision model with Predicate, Choice and Score modes.
- An unannounced GPT-6.1 Terra model surfaced in a public Microsoft code update that switched 18 evaluation configs to it before being reverted.
- Reuters reports OpenAI is speeding up a new fundraising round ahead of Anthropic's expected IPO.

ANTHROPIC 🔥:
- The Information reports Anthropic will release a new model this month that is especially good at reading users' underlying intentions, alongside Google's Gemini 4 Carbon and possibly a new OpenAI Terra model.

SPACEXSI 🔥:
- Grok Bot got a limited-time Pumpkin Bot Halloween avatar, and iPhone users can now launch their Bots with the Action Button.
- Runway MCP is now available in Grok Bot, bringing Runway generation into the chat.
- X rolled out a new AI ranking model for timeline recommendations plus a retrieval model trained to surface posts people read to the end.

META 🔥:
- Meta published a technical report on Muse Glimmer, its 30B Apache 2.0 open-weight agentic model that runs on a single consumer GPU.

MINIMAX 🔥:
- MiniMax officially launched M3.1 Flash, the former "Space Bunny" stealth model, at $0.15/$0.60 per 1M tokens, with open weights due Oct 23.

CLOUDFLARE 🔥:
- Cloudflare introduced Clairic, an extraction model that pulls, computes and forecasts data from unstructured inputs.

INCEPTION 🔥:
- Inception launched Mercury 3, a diffusion LLM that runs at over 1,000 tokens per second with a 1M-token context window.

FAL 🔥:
- fal and Black Forest Labs released FLUX 3 Max, a post-trained video model that generates a 10-second clip in about 12 seconds.

OpenAI prepares GPT-6.1 Terra for the upcoming release

OpenAI seems to be preparing to release GPT-6.1 Terra model as well.

> The model reference recently showed up in Microsoft's PR for evaluation configs.

> Earlier, The Information reported that this model is also in the works.

🗞 #openai @testingcatalog

Meta prepares Muse Confidential VM for private AI work

META 🔥: Confidential VMs feature is in the works on Muse web.

Meta previously teased that this feature will go live before the end of 2026 and will let users run Muse tasks in an encrypted, confidential environment.

According to a new hidden setting, Confidential VMs will initially be available only to Meta AI subscribers.

Voice mode, image and video generation, and wearables support won't be initially available if the user switches to Confidential VMs.

> It stays between you and your Muse - Your conversations run in a secure, encrypted environment.

> Your Muse data won't be used to improve AI at Meta.

> Some features aren't available yet - This includes voice mode, image and video generation, and wearables.

> Availability is limited. For now, Confidential VMs are only available to subscribers.

🗞 #meta @testingcatalog

OPENAI 🔥: Locked Chats in ChatGPT are rolling out to Free, Go, Plus, Pro, Business, Enterprise, and Edu plans.

Users can move chats into a protected space under a PIN on web and desktop, or a PIN, Face ID, Touch ID, or device unlock.

Locked in 👀👀👀

OpenAI ❤️ Orchestra Bio

> OpenAI has acquired Orchestra Bio⁠ to work on Rosalind models and tools for life sciences research.

> "Orchestra Bio is a platform that helps biopharma teams coordinate the complex process of drug development."

AI advancements in life sciences are probably one of the most desired areas for many. So far, it seems like a top priority for Anthropic, OpenAI, and Google.

PERPLEXITY 🔥: A fully local option is now available on the Perplexity Computer macOS Desktop app for Pro, Max, and Enterprise subscribers!

They did it 👀

Perplexity Computer no longer needs the internet or paid credits to work.

A fully local mode is powered by PPLX Pro, a post-trained version of the Qwen 3.8 model. For complex tasks, it can also run in Hybrid mode to use cloud models.

Perplexity Computer has a very good harness, and this needs a lot of testing time!

* 36 GB of unified memory is recommended.

MISTRAL 🔥: Voxtral 3, a new family of speech-to-text and text-to-speech models, is now available!

Voxtral Small Transcribe 3 landed in 3rd on the Artificial Analysis Speech-to-Text leaderboard, and Voxtral 3 TTS ranked in the top 12.

Voxtral 3 comes with 👀
- Accurate transcription across 50+ languages.
- Natural speech that adapts to the conversation.
- Human-like intelligence at spoken command.
- Pro Voice for distinctly accurate voice adaptation.

Voxtral 3 models are now available through Mistral's EU-hosted API or for self-hosting.

ICYMI 👀: ElevenLabs released Eleven Realtime Max and Mini speech-to-speech models.

Eleven Realtime v1 Max scores 98.8% on Big Bench Audio and starts speaking in 0.57s.

Eleven Realtime v1 mini is optimized for natural conversations at low cost.

Both models are available on ElevenAgents platform and APIs.

Perplexity Computer now runs fully locally on Mac

Perplexity’s Mac app now offers fully local Computer on Apple silicon, keeping model, agent, and task execution on-device. It can access private files and native apps with permission, giving users control over data and credit use.

🗞 #perplexity @testingcatalog

Mistral launches Voxtral 3 with a voice agent SDK

Mistral launched Voxtral 3 with speech-to-text, text-to-speech, and a real-time voice agent SDK. It supports hosted and private deployment, targets enterprise call workflows, and offers multilingual, low-latency voice processing.

🗞 #mistral @testingcatalog