OpenAI and Anthropic Are Racing to List First, ChatGPT Is Retiring Its Own Image Generator, and Claude Just Made a Price Cut Permanent

Three announcements this week had nothing to do with model capability and everything to do with how the two biggest labs are positioning for what comes next: who gets to Wall Street first, what "chatbot" even means once a flagship feature gets quietly retired, and whether a price cut is a promotion or a new floor. Here's what happened, and why each one is a bigger signal than it looks.

1. Anthropic Filed to Go Public First — and OpenAI Is Blinking

Anthropic filed a confidential S-1 with the SEC on June 1, targeting an October listing on Nasdaq at a valuation that could make it the first company ever to debut on public markets north of $1 trillion. Goldman Sachs, JPMorgan, and Morgan Stanley are lead underwriters, with a public S-1 expected around September and a roadshow running into early October. Anthropic's pitch leans on $47 billion in projected 2026 annual recurring revenue — roughly double OpenAI's — while OpenAI, which filed its own S-1 a week later on June 8, is now reportedly leaning toward pushing its listing into 2027 rather than list at anything below Sam Altman's stated floor of a $1 trillion valuation.

The subplot worth tracking isn't the valuation number, it's the sequencing. Both companies will need to raise tens of billions more in short order, and whoever prices first sets the comparable the other has to beat. Anthropic hiring Mariano-Florentino "Tino" Cuéllar as its first Chief Global Affairs Officer this month reads as part of the same push — a pre-IPO company building out the regulatory and government-relations bench a public company is expected to have, ahead of scrutiny that will only intensify once shares are trading.

2. ChatGPT Is Quietly Retiring the Feature That Made It Famous

OpenAI is retiring the official DALL·E GPT inside ChatGPT on August 30, telling users to download anything they want to keep before the cutoff. It's a narrower move than the headlines suggest — image generation in ChatGPT isn't going away, and user-built GPTs with image generation enabled are unaffected — but the preconfigured, named DALL·E surface, the one that put "type a prompt, get an image" into millions of people's hands starting in 2022, is being folded into the general-purpose ChatGPT Images pipeline. The same release cycle added Google Drive to ChatGPT's Library, letting users pull Docs, Sheets, and Slides directly into a conversation.

Put together, the pattern is consolidation: fewer named, single-purpose tools bolted onto the chat window, more capabilities absorbed into one default surface that reaches out to wherever a user's files already live. If your product still integrates against the DALL·E GPT specifically rather than the underlying image endpoint, that's a live migration, not a someday one:

Before Aug 30:  ChatGPT → "DALL·E" GPT (named tool call) → image
After Aug 30:   ChatGPT → ChatGPT Images (default generation path) → image

Anything hardcoded to the named GPT — deep links, saved prompts, browser automations — should be pointed at the general image-generation flow before the 30th.

3. Anthropic Turned a Sunset Clause Into a Floor

Claude Sonnet 5 launched in June at introductory pricing of $2 per million input tokens and $10 per million output tokens, with a September 1 reversion to standard rates of $3/$15 already on the calendar. On August 11, Anthropic reversed that plan and made the $2/$10 rate permanent instead. For any team that built a cost model around the scheduled 50% increase, this is good news that still deserves a second look — a lab willing to cancel a planned price hike in a competitive market is also a lab that can revisit pricing again with less notice next time.

It's also a data point in the same story as the IPO race: locking in aggressive API pricing ahead of a public listing is a bet that volume and lock-in matter more right now than near-term margin — the same logic Google and OpenAI have leaned on with their own price cuts this year.

What Connects the Three

None of this week's news is about a smarter model. It's about the infrastructure around the model: which lab gets to Wall Street on better terms, which single-purpose features get absorbed into a general default as the product matures, and which pricing commitments a lab is willing to make permanent to keep developers building on top of it instead of a competitor. If you're choosing a foundation model provider for a product with a multi-year horizon, this is the week to stress-test your assumptions against all three — not just benchmark scores, but who's about to answer to public shareholders, what "stable" API surface actually means, and whether your unit economics survive a rate that isn't guaranteed to stay introductory forever.

— Maya