OpenAI COOKED

Summary
Matthew Berman recaps the major product and model announcements from OpenAI DevDay 2026. He reviews the launch of "dots" always-on personal agents, the Cerebras-powered Ultrafast generation tier, the new Pro 500 tier, GPT-6.1 Sol, Codex Cloud updates, Decisions API, and ChatGPT Space.

What is shown

• [00:10] The physical ModRetro handheld console given out to DevDay attendees for building retro games with Codex.
• [00:17] OpenAI's promotional video introducing "dots" autonomous agents powered by GPT-6 Astra, operating within a dedicated cloud computer interface.
• [00:40] Official overview text detailing dots' 24/7 background operation, team integrations (Slack, Teams), and connections to 4,000+ app plugins.
• [01:11] UI walkthrough of dots operating apps (Blossom Music) in a browser desktop environment, comparing the mascot design language to GrokBot and Muse.
• [02:23] Official access details for dots in ChatGPT Pro and Business tiers, including usage quota guidelines.
• [03:06] Side-by-side demo comparison clip between Standard generation and Ultrafast generation assembling and launching a 3D rocket model.
• [04:40] OpenAI pricing update announcement detailing the new $500/month Pro 500 subscription plan and the revised Pro 200 plan limits.
• [05:03] Announcement video and API pricing comparison table for GPT-6.1 Sol, GPT-6 Astra, and GPT-6 Luna.
• [06:10] Evaluation graphs comparing GPT-6.1 Sol against GPT-6 Sol and GPT-6 Astra across DeepSWE, GDP.pdf [06:58], and OSWorld 2.0 [07:08].
• [07:21] Codex Security Cloud dashboard demonstrating continuous GitHub repository scanning with Daybreak Blue.
• [07:35] Codex in the Cloud mobile and desktop interfaces, refreshed CLI, and desktop Code Review tool.
• [08:17] Overview slide of Decisions API, showcasing real-time routing and classification based on GPT-6 Luna.
• [08:49] Plugin extensions platform demo showing third-party web apps (Canva) embedded natively inside ChatGPT.
• [09:38] ChatGPT Space collaborative workspace UI featuring shared team documents, boards, and integrated agents.

Claims & numbers

• The presenter states DevDay attendees received a physical ModRetro handheld device to create custom games using Codex [00:10].
• The presenter notes dots are always-on agents powered by GPT-6 Astra that can connect to over 4,000 plugins, Slack, and Microsoft Teams [00:17, 00:40].
• Ultrafast tier uses Cerebras wafer-scale chips to deliver up to 8x faster token generation (300 tokens/sec in Codex and up to 6x faster in the API) compared to standard Nvidia GPU infrastructure [03:11].
• The presenter states Ultrafast costs 6x more than standard generation, claiming he burned through $1,000 worth of tokens in approximately 1.5 hours in early testing [03:26, 03:31].
• OpenAI launched a Pro 500 subscription plan priced at $500/month featuring 25x ChatGPT Plus usage limits and Ultrafast access [04:44].
• The presenter claims the $200/month Pro 200 plan's limits were cut from 20x Plus allowance down to 10x [04:49].
• Pricing for GPT-6.1 Sol is $2.00 per million input tokens, $10.00 per million output tokens, and $0.10 cached input, compared to GPT-6 Astra ($10.00 input / $50.00 output) and GPT-6 Luna ($0.10 input / $0.50 output / $0.01 cached) [05:50].
• Benchmarks show GPT-6.1 Sol matching or exceeding GPT-6 Astra on DeepSWE and GDP.pdf, and nearly matching it on OSWorld 2.0 at one-fifth the cost [06:10–07:19].
• The Decisions API is built on GPT-6 Luna's lightweight model architecture for fast agentic routing and decisions [08:29].

Notable quotes

• [03:11] "This is GPT-6 Astra powered by Cerebras chips, and it goes eight times faster than what Astra would do normally on their Nvidia GPUs."
• [04:56] "And what that means is you're barely getting more than what you used to get at the $200 mark, but now you're going to be paying a lot more."
• [05:41] "GPT-6.1 Sol is a phenomenal model. It is basically as good as Astra and much cheaper and faster."

Assessment
This is an independent creator review and recap of official OpenAI DevDay 2026 announcements. The presenter displays OpenAI's official announcement tweets, promotional video clips, and benchmark charts, mixing his personal testing anecdotes with official marketing claims without performing live benchmarks on-screen.

Described by gemini-3.8-flash on 2026-09-30 from the video's audio and frames.