OpenAI just made its $200 Pro plan worse. That’s the part most people will miss while reacting to the new $500 tier.
As reported by Engadget, the announcements came out of OpenAI’s DevDay event. The new Pro 500 plan is positioned as the highest usage tier the company offers, and its headline feature is something called “Ultrafast,” a premium speed tier that pushes Codex and Work up to 300 tokens per second. OpenAI describes it as built for “workloads where speed matters most.” Ultrafast access is also available to API users, so this isn’t just a consumer play.
But the $500 plan didn’t arrive alone. At the same time, OpenAI reduced what $200 Pro subscribers actually get. In Codex and Work, included usage drops from 20x the Plus allowance down to 10x. In ChatGPT, GPT-6 Pro message caps fall from 200 to 100 per week. OpenAI says existing subscribers will keep their current limits temporarily, then receive a one-time credit to soften the transition. That’s a meaningful cut, not a minor adjustment.
This is a familiar move in subscription businesses: introduce a higher tier, then reposition the tier below it as the mid-range option rather than the premium one. The $200 plan, which OpenAI introduced at the end of 2024 as the first of its kind in the industry, quickly got matched by Anthropic, Perplexity, and others. Now OpenAI is trying to pull away again on price. Whether competitors follow to $500 is worth watching closely.
There are a few other additions bundled in. All Pro subscribers, at either price point, will receive one “Dot,” which is what OpenAI is calling its personal AI agents. Conversations with your Dot won’t count against your ChatGPT usage limit. OpenAI is also rolling out a “Sign in with ChatGPT” option in partnership with 16 companies including Notion and Cognition, letting Plus and Pro customers run OpenAI models directly inside those products without needing to configure an API key. That’s a meaningful friction reduction for non-technical users and small teams.
The other announcement from DevDay is GPT-6.1 Sol, a new model OpenAI says delivers performance close to its current frontier system, GPT-6.0 Astra, in coding and computer use tasks, at one fifth of the cost. OpenAI also claims it’s safer, citing lower failure rates on transparency about tool limitations, respecting user-set restrictions, and avoiding unauthorized actions in agentic contexts. That last point matters more as these models get used for longer, more autonomous workflows where a model quietly doing something it shouldn’t is a real problem.
- GPT-6.1 Sol is available today in Codex and Work for Plus, Pro, Business, Enterprise, and Education users
- ChatGPT availability for GPT-6.1 Sol follows at a later date
- Ultrafast speed tier is available to both Pro 500 subscribers and API users
- “Sign in with ChatGPT” launches with 16 partner integrations including Notion and Cognition
So what’s the net effect here? OpenAI is segmenting its user base more aggressively, pushing power users toward $500 while making the $200 tier look more like a mid-market product. For developers and teams evaluating where to spend, the GPT-6.1 Sol pricing story is probably more interesting than the subscription tiers. A model approaching frontier performance at 20% of the cost is a real change in the cost structure of building with OpenAI. That’s the part worth paying attention to.
