OpenAI Unveils Always-On Dots Agents, Boosts GPT-6.1 Sol Performance at DevDay
OpenAI's DevDay pushed AI agents and model upgrades, while Anthropic's leaked IPO prospectus revealed a $2 trillion target amid broader industry safety discussions.
The story
The AI software landscape saw a flurry of activity, with OpenAI's DevDay 2026 taking center stage on September 29, and its implications being discussed this week. The company unveiled 'Dots,' always-on AI agents powered by GPT-6 Astra, designed for persistent automation across thousands of applications.
This move signals a deeper push into autonomous AI systems. OpenAI also announced GPT-6.1 Sol, an upgraded model that reportedly delivers near-Astra level performance at one-fifth the cost, aimed at agentic coding and professional tasks.
However, the event was also marked by a critical safety signal: OpenAI cancelled the planned release of GPT-6.1 Astra due to internal safety concerns, highlighting the growing challenges in deploying advanced AI. Separately, Anthropic's confidential IPO prospectus leaked this week, revealing a staggering $2 trillion valuation target, alongside a $42 billion net loss for 2025 against $4.6 billion in revenue. Meanwhile, the U.S. administration announced a voluntary AI safety accord with major labs, including OpenAI, Anthropic, Meta, and Google, focusing on internal security and external audits.
Who moved
OpenAI
What Changed: The company launched 'Dots,' a system of always-on AI agents, alongside an upgraded GPT-6.1 Sol model and a new $500/month Pro 500 subscription plan at its DevDay 2026 conference.
Consequence: This expands OpenAI's product offerings into persistent, autonomous automation and provides developers with a more cost-effective, high-performing model, while the Pro 500 plan targets power users with increased usage and speed.
Anthropic
What Changed: Its confidential IPO prospectus was leaked, revealing a target valuation of $2 trillion and disclosing a $42 billion net loss in 2025 against $4.6 billion in revenue.
Consequence: The leak provides an unprecedented look into the financial ambitions and current operational costs of a major AI lab as it eyes a public listing, underscoring the high capital intensity of the AI race.
US Government
What Changed: It announced a voluntary AI safety accord with several large AI companies, including OpenAI, Anthropic, Meta, and Google.
Consequence: This accord aims to strengthen internal security and promote collaboration with external auditors, signaling a collective industry effort towards safer AI development, though without immediate penalties for non-compliance.
Google DeepMind
What Changed: Its Gemini 4 Argon model is being rolled out cautiously, with initial access limited to selected cybersecurity partners rather than a wide public release.
Consequence: This indicates a measured approach to deploying powerful new models, prioritizing specific, high-stakes applications and controlled environments for advanced reasoning capabilities.
New models
Dots
Lab: OpenAI
What: Always-on personal AI agents powered by GPT-6 Astra.
Use: For persistent automation across thousands of applications and user workflows.
GPT-6.1 Sol
Lab: OpenAI
What: A major upgrade to GPT-6 Sol, offering near-Astra performance at one-fifth of Astra's token price.
Use: Designed for agentic coding, computer use, and professional work, providing a more economical option for complex tasks.
Market signals
Anthropic's leaked IPO prospectus targets a $2 trillion valuation, while disclosing a $42 billion net loss for 2025 against $4.6 billion in revenue.
Implication: This highlights the immense investor confidence in future AI growth, coupled with the significant capital expenditure and losses currently required to build and scale frontier AI models.
AI startup Instinct's valuation quadrupled to $10 billion in one month.
Implication: This rapid repricing illustrates intense investor demand and the accelerated pace of valuation growth in the AI agent sector, potentially outpacing traditional diligence timelines.
Cybersecurity firms Island and Cyera each secured $400 million funding rounds.
Implication: These substantial investments reflect the growing need for AI-powered security solutions, particularly to protect the burgeoning ecosystem of AI agents operating within enterprises.
What we'll be watching
- OpenAI's DevDay Exchange events will continue in cities across Asia, Europe, and Latin America throughout October and November 2026.
- OpenAI plans a broad release of its Decisions API in the coming days, following its limited preview at DevDay.
- An Ultrafast variant for OpenAI's GPT-6.1 Sol model is expected to follow, offering premium inference speed.
- Anthropic is reportedly aiming for its initial public offering (IPO) as soon as October 2026.
Reporting + analyst voices: grounded via Google Search at publish time.