OpenAI releases GPT-6 model to more ChatGPT users

#1 📝 OpenAI News

GPT-6 and Intelligent UI for everyone - OpenAI News reports that OpenAI released a new GPT-6 model on October 7, 2026, making it available to more people after introducing the first GPT-6 models for paid customers the previous month. The new model was built for the more than 1.2 billion people who use ChatGPT each week.

Also covered by: @OpenAI News, @OpenAI, @Sam Altman

#2 📝 Anthropic News

Introducing Claude Haiku 5.5 - Anthropic announced Claude Haiku 5.5 in an October 7, 2026 post.

Also covered by: @Cognition, @Cursor, @Claude, @Claude, @Mike Krieger, @Cursor

#3 𝕏

Aravind Srinivas announced the open-sourcing of pplx-embed-v2-late, 9B and 0.6B multi-vector models for text and images in one shared embedding space, with weights available on Hugging Face. It supports multimodal indexing with 9B, on-device querying with 0.6B, and PDF search without OCR, scoring 92.4% on MADQA and 64% on BrowseComp+.

#4 𝕏

LlamaIndex 🦙 announced OpenDocRouter, a live API for switching among 10 frontier and open-source document-parsing models in one line while keeping the same request and Markdown output. Models are scored on ParseBench for quality and cost, with optional grounded layout data and per-token pricing from $0.86 to $48.82 per 1,000 pages; failed pages aren’t charged.

#5 𝕏

DeepLearning.AI recapped Google’s Gemini 3.8 Live models, which listen, reason, and respond in one system while accepting image and video input. The Extended Thinking version ranks first on Artificial Analysis’ Speech to Speech Index, while the standard version ranks second in blind live conversations judged by people and costs $0.84 per hour of input audio—the index’s lowest price.

#6 𝕏

Harrison Chase announced that deepagents now supports dynamically loading tools when a skill is loaded, pairing skill-specific tools without making them always available. With OpenAI or Anthropic models, this can be done without breaking the prompt cache.

#7 𝕏

DeepLearning.AI recapped how DeepSeek reduced agent memory, reporting 890 bytes of cache per token—437x smaller than DeepSeek-V1—and +25% compute per output token from 4K to 1M input. It also reported Flash beat V4-Pro on the AA Index, 39 vs 36, at $0.27 vs $0.67 per task.

#8 𝕏

claire vo 🖤 shared an x.ai bot that Zach Davis made, which reviews the @cxodev website and flags small design bugs and unusual issues accidentally shipped. The post also linked to x.ai, where others can get their own version.

#9 𝕏

Rowan Cheung said he built an app with GPT-6 Astra that uses live weather and photos of him to pick his outfit each morning. Setup involves a few full-body photos, one face photo, measurements and fit notes, color preferences, and a master prompt, with 1 to 2 hours of build time.

#10 𝕏

NVIDIA AI created PivotOPD to teach AI agents to avoid and recover from mistakes. During training, a teacher model demonstrates a better action and how to get back on track over the next few steps.

#11 in

Dan Shipper recapped OpenAI’s Dots, saying they launched the previous week. He said his Dot posted model-testing results to Slack and brought back replies; people at Every Inc. use Dots to extract information from school emails, catch missed Slack messages, and flag access requests; and head of video Randy Counsman used voice to guide his Dot through video edits. Shipper called Dots roughly on par with Muse, Instinct, and Grok bots, but said permissions can require repeated approvals and that he expects to use a persistent agent within a year.

#12 in

Udi Menkes recapped ten takeaways from Higgsfield AI Co-Founder & CEO Alex Mashrabov’s appearance on 20VC with Harry Stebbings, stating that Higgsfield is “No. 3 from $1M to $1B,” “$99 a month became a $6M+ annual deal,” and the company grows without paid acquisition, owns distribution, and spends more than $4M a month on models. Menkes also highlighted eight creative directors steering Higgsfield toward camera control, proprietary models driven by customer demand, 80%+ margins from open and owned models, Asian DTC brands rebuilding GTM around ad volume, and a $10B+ short-form drama market going AI end to end, while revenue points West and usage points to Seoul.

Also covered by: @Udi Menkes 🚢, @Udi Menkes 🚢

#13 in

Udi Menkes recapped how Sierra founders Bret Taylor and Clay Bavor are developing the open Personal Agent Protocol (PAP) with Meta, spanning four layers: discovery, OAuth authentication, action, and permissions. He argues builders should design for two customers—the person and their agent—and said specification version 0.1 was expected later that month.

#14 𝕏

Mistral AI shared that Mistral Large 4 was evaluated on 657 business workflows in AutomationBench using simulated Gmail, Google Sheets, Slack, and Salesforce apps, finishing ahead of Kimi K3, MiMo-V2.6-Pro, and DeepSeek V4 Pro.

#15 𝕏

Mistral AI shared that Mistral Large 4 is the #1 oss model on Harvey’s Legal Agent Benchmark (LAB).

#16 𝕏

Google DeepMind says SynthID Detector is now available to everyone, enabling users to check whether online content was generated using Google AI or tools from OpenAI, NVIDIA, and Kakao. Support for content generated with Apple tools is coming soon.

Also covered by: @NVIDIA AI, @Google AI, @Google DeepMind

#17 𝕏

Sam Altman shared OpenAI’s “Sharing AI progress in mathematics” webpage, stating, “We are entering a new era of discovery now.”

#18 𝕏

OpenAI shared progress on ChatGPT for Teens, which automatically applies protections by default to accounts identified as belonging to someone under 18. It also previewed College Planner, a coming-soon tool that will organize application requirements, deadlines, tasks, and financial-aid steps for U.S. high school students planning to attend a four-year college.

#19 in

Claire Vo quoted Zach Davis arguing that problems encountered with AI are users’ responsibility, not AI’s. She added that CXO.dev views repo readiness as foundational to agent and human effectiveness and advocates investing earlier and more heavily in developer platforms.

#20 𝕏

Boris Cherny says he has Claude direct-message people on Slack, monitor their responses, and create a chart with the results.