OpenAI is finally delivering on the hype that it attempted to build over the past few weeks for its DevDay 2026, replete with over 20 distinct product launches, albeit complicated by an evolving competitive environment where Anthropic has again taken an apparent lead by launching the eminently stellar Opus 5.5 and Sonnet 5.5 AI models.
Here’s everything you need to know before tuning into OpenAI’s DevDay 2026 event
OpenAI has suffered a series of setbacks recently. First, Anthropic launched its Claude Opus 5.5 model last week that not only exceeded the capabilities of its Fable 5.1 model but also beat OpenAI’s reigning apex model, GPT-6 Astra, on most benchmarks. In fact, Claude Opus 5.5 on a medium intensity beat GPT-6 Astra on max for a fifth of the cost per task on OSWorld 2.0 Offline.
OpenAI did try to put up a fight by debuting GPT-6 Sol and GPT-6 Luna models. However, as we had noted at the time, while GPT-6 Luna is cheaper than even the off-peak API pricing for the DeepSeek V4.1 Flash model, which offers input at $0.15 per 1 million tokens and output at $0.60 per 1 million tokens, DeepSeek V4.1 Flash still scored 74.2 on DeepSWE v1.1 vs. GPT-6 Luna’s score of around 67. What’s more, Anthropic’s Claude Opus 5.5 is currently leading the scoreboard when it comes to the Artificial Analysis Intelligence Index with a score of 58 vs. 48 for GPT-6 Sol.
Additionally, on Monday, Anthropic released the newest iteration of its workhorse model, Claude Sonnet 5.5, which scored 56 on Artificial Analysis’ Intelligence Index, placing it behind Claude Opus 5.5 but above GPT-6 Astra and miles above GPT-6 Sol. What’s more, according to Anthropic, the Sonnet 5.5 is around 30 percent faster and cheaper than the Sonnet 5.0 on most tasks even though it is priced at the same $2 per 1 million tokens of input and $12 per 1 million tokens of output.
And to make matters worse, the Wall Street Journal reported on Monday that OpenAI has apparently scrapped its GPT-6.1 Astra model after it showed a higher proclivity for engaging in deception, while breaking regression testing, hallucinating tool calls, ignoring user guardrails, and failing to follow basic instructions.
And, on the consumer front, Meta’s new personal AI agent, Muse, has been breaking adoption records, far outpacing the initial uptake of ChatGPT and Gemini apps.
In what is a rare bit of positive news, however, Axios has just reported that OpenAI’s annual recurring revenue (ARR) is now close to $70 billion ($68 billion to be exact) after growing by over 70 percent since the beginning of Q3 2026. For reference, Anthropic’s ARR hovered near $65 billion in August.
As for what you can expect during Sam Altman’s keynote address scheduled for 10:00 a.m. PT, OpenAI will likely unveil personal AI agents called Dots. According to the leaker M1, “you can text it, call it, Slack it, and email it, and it can buy things for you with your approval.” Each Dot also gets its own 3D avatar that you can design.
Do note that OpenAI had teased an always-on agent, called “o,” over the weekend. It is unclear if Dot and “o” are the same product or represent entirely different launches.
Other expected announcements include:
- Codex plugins
- Multiplayer Codex
- Details on a 1,000-hour internal agent run
- The $200 Pro subscription tier is expected to reopen sans 5-hour usage cap, but will only offer around half of the previous iteration’s bang for the proverbial buck.
- Collaborative Agent Spaces
OpenAI’s DevDay 2026 goes live
You can watch Altman’s keynote address here. We’ll continue to update this post with pertinent details. So, stay tuned!
Meet Dots: Your personal, always-on agents
As expected, OpenAI has just unveiled Dots, your personal, always-on AI agents powered by GPT-6 Astra. With support for OpenAI’s plugins, Dots can connect to over 4,000 apps, including ChatGPT, Slack, and Teams. You can interact with Dots via text or hop onto a call.
OpenAI has given each Dot access to its own cloud computer (VM) and browser. When combined with the access to a plethora of apps via plugins, Dots become a veritable productivity powerhouse. According to OpenAI, you can message or call Dots in ChatGPT on desktop, web, and mobile. They can also message you with progress, questions, or decisions that need your input. You can also message your Dot in Slack and Teams, with support for texting expected to land soon.
As for built-in safeguards, OpenAI notes:
“Each Dot works on its own cloud computer, while your computer and its contents stay separate unless you choose to connect it. For signing into supported websites, Dots can use saved passwords without exposing them to the model.”
OpenAI also assures that it “does not use content from ChatGPT Business, Enterprise, or Edu workspaces to improve our models by default,” while those holding personal ChatGPT plans can opt out at anytime.
Companies can also setup specialist Dots, replete with their own identity, credentials, and access to the systems they need to complete their tasks.
In what is a big deal, OpenAI says that it is also bringing Dots to Microsoft Agent 365, allowing businesses to manage these agents via tools that they already use. Also, the AI lab is partnering with Amazon to host Dots on AWS.
Finally, OpenAI says that Dots are rolling out today in ChatGPT to Pro and Business Premium users in all markets excluding the European Economic Area, Switzerland, and the UK. Enterprise users (including Edu and Healthcare) can try the beta when their workspace admin enables it. Also, your first Dot is included in your Pro or Business Premium plan at no extra cost. Moreover, for the next month, your Dots usage won’t count toward your plan’s allowances (conversations only and not actual tasks).
We should note though that Dot’s debut was not without a hitch, with the agent failing to respond during OpenAI’s demo. Also, we can’t but help notice that Dots appear to be geared more towards enterprises, which should give Meta’s Muse a relatively unhindered path for now to dominate the consumer-oriented personal AI agent space. The fact that the Dot is only available on the Pro tiers while Meta Muse is free up to a limit will also skew the uptake.
GPT-6.1 Sol
OpenAI has just announced GPT-6.1 Sol, its new workhorse model that supposedly offers “near-Astra level intelligence at a fifth of the price” on DeepSWE v1.1, which evaluates complex software-engineering tasks in real codebases. According to Altman, “the cache input costs [$0.10 per 1 million tokens] are a 95% discount on standard input. So this makes it much cheaper for agents to reuse context, iterate on solutions, and work through longer tasks.”
According to OpenAI, the model scores higher than Anthropic’s Opus 5.5 on GDP.pdf, which measures how accurately models answer professional questions using complex PDF documents, including tables, charts, diagrams, and fine-print details. It also approaches GPT‑6 Astra’s performance on the benchmark at roughly one-fifth the cost per task.
On AutomationBench, which measures whether agents correctly complete multi-step business workflows, GPT‑6.1 Sol scores 2.2 percent above Opus 5.5’s score at medium reasoning effort, at roughly a third of the cost.
On OSWorld 2.0’s offline set, which evaluates agents on demanding computer-use workflows, GPT‑6.1 Sol comes within 2.1 percentage points of Astra’s score at maximum reasoning effort at roughly one-seventh the cost per task.
According to OpenAI, GPT‑6.1 Sol also improves factual accuracy on difficult prompts, with the largest factuality improvement over GPT‑6 Sol coming on low reasoning effort, where it reduces the share of responses containing a factual error from 11.4 percent to 7.7 percent, with the overall error rate remaining within 1.9 percentage points of GPT‑6 Astra’s, at less than one-fifth the cost per task.
The model’s standard API prices are $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens.
And, if you were wondering why OpenAI would launch a new Sol model just a few days after the GPT-6 Sol, it is intended to paper over the embarrassment caused last week.
Ultrafast, OpenAI’s new premium speed tier for workloads, and Private Intelligence
OpenAI’s Ultrafast offers up to 8× faster token generation (300 tokens per second) in Codex and up to 6x in the API. Leaked tidbits in recent days had indicated that this tier would be powered by Cerebras’ Wafer-Scale Engine, but OpenAI has yet to confirm it.
GPT‑6 Astra Ultrafast is available today in the API and in ChatGPT Work and Codex on Pro 500 and Enterprise plans, while GPT‑6.1 Sol Ultrafast is expected to land soon. It costs an insane $60 per 1 million tokens of input and $300 per 1 million tokens of output though!
Moreover, OpenAI says that its Zero Data Retention with Private Safety Processing (ZDR with PSP) enables offline, automated safety review without the AI lab retaining any customer prompts, responses, or underlying content.
OpenAI is also about to launch Private Inference this fall, which combines confidential computing with strict, verifiable controls.
Other announcements
Here are some of the other major announcements from OpenAI today:
- Codex in the cloud, enabling access via a computer, remotely from a phone, or in the cloud from any device.
- The Agents API now supports computer use so developers can build agents that interact with software to complete tasks.
- Plugin extensions let you give your plugin a home in the sidebar and build interactive panels where people can work alongside the conversation.
- ChatGPT Space, a dedicated space where teammates, ChatGPT, and your Dot agent can build on shared knowledge and data.
- Pages, which launch in Q4 2026, are a new type of document, built for human and agent collaboration. Anything you can do in ChatGPT you can also do on a page: write, research, generate charts, create images, or visualize information.
- OpenAI plan allowances are now shared across 16 partners, including Cognition’s Devin, Notion, Vercel, T3, OpenClaw, and Dactyl.
- OpenAI has just launched a new Pro tier, called Pro 500, which offers 25x the allowance of the ChatGPT Plus tier. Starting at $500 per month, the tier gives access to GPT-6 Astra Ultrafast, as well as Pro-level reasoning powered by Astra, Codex, deep research, image creation, memory and file uploads, and the Dots agents. In what is a critical clarification though, Dots will be accessible via all Pro-tier plans, including the Pro 100 plan.
- The AI lab has also just reset all plan allowances to give you a fresh slate.
About the author: Writing is my one incontrovertible passion. Over the past six years, he has authored over 2,200 distinct articles on financial and tech-related topics, spanning nearly 1 million words. And he has been a member of Wcctech mobile team since 2025.
As an alumnus of the University of Toronto, Rotman Commerce Program, I bring nuance, in-depth knowledge, and a unique perspective to every topic that I cover.
When I’m not writing, I’m traveling the world, exploring hidden confectionaries and restaurants as an aspiring food connoisseur.
Follow Wccftech on Google to get more of our news coverage in your feeds.
