Skip to main content
AI Learning Resources

Quick Overview

This video is an episode of The Code Report presented by Fireship. It covers the major announcements from OpenAI's DevDay keynote, examining new models, developer tools, and pricing changes across the AI ecosystem.

Key Points

  • 1.OpenAI introduced Dots, a subscription-based autonomous AI agent feature powered by GPT-6 Astra, following their hiring of OpenClaw creator Peter Steinberger.
  • 2.A new Pro 500 subscription tier priced at $500 per month adds Ultrafast inference speeds of up to 300 tokens per second, while the existing Pro 200 plan saw its compute allocation halved.
  • 3.OpenAI unveiled GPT-6.1 Sol as a cost-efficient model providing near-Astra intelligence at two dollars per million input tokens and ten dollars per million output tokens.
  • 4.The new Decisions API uses the lightweight Luna model to perform rapid classification from predefined choice sets without parsing free-form JSON strings.
  • 5.Sign in with ChatGPT allows third-party developer platforms to authenticate users and consume the user's existing subscription token allowance directly.

Summary

The presentation opens with a look at the escalating competition in personal AI assistants, highlighting Meta's announcement of Muse and OpenAI's subsequent unveiling of Dots at DevDay in San Francisco. Dots originated from the open-source assistant project Clawdbot, later renamed OpenClaw, created by developer Peter Steinberger before he was hired by OpenAI. Configured as always-on personal agents powered by GPT-6 Astra, Dots were shown managing daily engineering tasks. In a keynote demonstration, Sam Altman commanded a dot named Alfred to migrate code away from a legacy inventory API, resulting in automated dependency tracing, test execution, and three pull requests submitted directly to GitHub.

OpenAI outlined new subscription tiers to support these agentic capabilities, starting with the Pro 500 plan priced at $500 per month. The tier offers 25 times the usage of standard ChatGPT Plus and includes access to Ultrafast mode, which outputs GPT-6 Astra responses at 300 tokens per second. In the API, Ultrafast context is billed at $60 per million input tokens and $300 per million output tokens. Concurrently, the existing $200 per month Pro tier was reopened with its compute allocation reduced by 50 percent. For budget-conscious developers, OpenAI introduced GPT-6.1 Sol, priced at $2 per million input tokens and $10 per million output tokens, matching the cost of GPT-6 Sol while delivering intelligence close to GPT-6 Astra.

To improve reliability in agentic workflows, OpenAI introduced the Decisions API. Rather than prompting a large language model to produce free-form JSON that requires defensive parsing and retry logic, the Decisions API uses the lightweight Luna model to evaluate a prompt against a strictly defined array of acceptable choices. The model selects the appropriate answer in milliseconds, ensuring valid typing by design and preventing hallucinations. This approach mirrors the system architecture of Diogo Almeida's TypeSafe AI and its Jev model.

Additional announcements included Pages, Space, a Dots integration for Slack, and the developer-focused Sign in with ChatGPT authentication service. Supported by 16 launch partners including Devin, Notion, Vercel, and Lovable, the login system allows third-party tools to consume the end user's ChatGPT plan allowance instead of requiring developers to front API costs. The video concludes with an overview of open-weight decision alternatives from Fastino Labs, including GLiNER2.5-Decide and GLIDE, which enable developers to run fast local classification and intent routing without vendor lock-in.

Introduction of Dots and OpenClaw Background

The report recaps the origins of OpenAI's new Dots assistant feature, tracing it back to Peter Steinberger's open-source OpenClaw project and Meta's competing Muse assistant. OpenAI demonstrated Dots acting as autonomous agents capable of performing repository migrations, opening pull requests, and handling scheduled tasks under a dedicated hundred-dollar monthly Pro plan.

Pro Subscription Updates and Ultrafast Tier

OpenAI detailed adjustments to its developer and individual pricing tiers, highlighted by the introduction of the Pro 500 plan at five hundred dollars per month. This tier provides twenty-five times standard Plus usage and unlocks Ultrafast mode, delivering GPT-6 Astra throughput at three hundred tokens per second alongside API pricing of sixty dollars per million input tokens and three hundred dollars per million output tokens.

GPT-6.1 Sol and the Decisions API

The presentation covered the release of GPT-6.1 Sol, which targets complex workloads at a fraction of flagship pricing. OpenAI also announced the Decisions API powered by the Luna model, offering constrained classification against fixed sets of options to eliminate latency and formatting errors associated with generating and parsing full JSON text responses.

Sign in with ChatGPT and Open-Weight Decision Models

OpenAI rolled out Sign in with ChatGPT with sixteen launch partners, enabling developers to charge API usage directly to end-user subscriptions. The video contrasts proprietary decision APIs with open-weight alternatives such as Fastino Labs' GLiNER2.5-Decide and GLIDE models, which allow developers to self-host task-routing and classification logic locally.

The Bottom Line

The video provides an overview of OpenAI's DevDay announcements, focusing on the commercialization of autonomous agents, adjusted subscription tiers, and optimized classification APIs. It establishes how features like Sign in with ChatGPT and the Decisions API alter the economics of building AI applications while highlighting the trade-offs of relying on proprietary platforms. What remains unresolved is how effectively developers will adopt closed ecosystem tools compared to self-hosted open-weight alternatives.

FAQ

What is OpenAI DevDay and what are the major product features announced during the event?

OpenAI DevDay is a developer conference where OpenAI unveiled several new tools, including Dots autonomous agents, the high-speed Ultrafast inference tier, GPT-6.1 Sol, the Decisions API, and Sign in with ChatGPT.

What is the OpenAI Decisions API and how does it handle structured outputs?

The Decisions API uses OpenAI's Luna model to classify input against a predetermined list of allowed answers, returning a structured selection within milliseconds without requiring JSON parsing or retry loops.

How does the Sign in with ChatGPT feature benefit third-party software developers?

Sign in with ChatGPT allows developers to authenticate users in their applications and draw token consumption directly from the user's existing ChatGPT subscription rather than paying for API usage themselves.

What are the token limits and pricing details for the OpenAI Pro 500 subscription plan?

The Pro 500 plan costs $500 per month, provides 25 times the usage limits of ChatGPT Plus, and includes access to the Ultrafast model tier running at 300 tokens per second.

How does Fastino Labs GLiNER2.5-Decide compare to proprietary decision-making AI models?

GLiNER2.5-Decide is an open-weight model that runs locally on developer hardware up to eight times faster than cloud-only decision APIs while giving developers full ownership of the model weights.

Worth watching for

Software engineers, AI developers, and tech entrepreneurs interested in OpenAI's latest product announcements, subscription changes, and structured classification APIs.

  • openai
  • chatgpt
  • ai-agents
  • devday
  • machine-learning
  • fastino
Watch on YouTubeMore videos from Fireship