Google Nears Gemini 4 Release 12 Days After Amodei's Call to Slow Down

On September 12, Anthropic CEO Dario Amodei published an essay titled "We Must Pace the Frontier." Ten frontier labs signed on to the coordination framework. The trade press read it as a commitment to slow down release schedules.

Twelve days later, the information reported Google is nearing release of Gemini 4, its flagship model. This report came 48 hours after Anthropic shipped Claude Opus 5.5 at 20 percent lower pricing and OpenAI shipped GPT-6 Sol at half of GPT-5.6 Sol pricing plus GPT-6 Luna at a quarter of its predecessor’s price.

That’s three frontier labs moving in 12 days, and the framework held back none of them.

What Wednesday's piece missed

Business of GTM's Wednesday tent-pole “The 10-day slowdown” argued that the frontier-lab pacing narrative had collapsed within 10 days, using evidence from Anthropic and OpenAI. That piece left Google as an open question, since Google hadn’t signed the framework but was treated as implicitly part of it. Thursday's news answers that question.

Google's Gemini 4 arriving in the same 12-day window also sharpens the legal question for the antitrust plaintiffs: was this coordination or competition? If the September 12 essay was a coordination signal, three of the largest frontier labs releasing models within 12 days gives the plaintiffs the discovery calendar they wanted. If the essay was a policy pitch framed as coordination, three labs releasing anyway is evidence against the plaintiffs. Either way, the pacing never happened, and the two readings differ only on whether the labs face legal exposure for saying it would.

The three-lab price floor is the enterprise buyer's new anchor

For GTM teams, the bigger story is the enterprise buyer's Q4 procurement conversation.

Twelve days ago, most enterprise AI buyers compared frontier-lab pricing across two vendors, Claude and GPT. Now there are three. Claude Opus 5.5 costs $4 per million input tokens and $20 per million output tokens. OpenAI GPT-6 Sol costs $2 and $10, and GPT-6 Luna at $0.10 and $0.50. Gemini 4 pricing comes next.

That changes the negotiation for every enterprise AI contract in the pipeline. When buyers compare two vendors, each can argue it’s the safe pick. With three, each vendor also has to answer the price pressure from a lab that has historically priced aggressively against the other two.

AI-native SaaS companies modeling 2027 unit economics on frontier-lab API prices now have a third price point to plan around, and it will probably be low. Google has historically priced Gemini below Claude and GPT on both input and output tokens at comparable capability tiers, and it rarely launches a flagship against Anthropic or OpenAI without an aggressive pricing hook.

Google announced Gemini 4 Argon on September 30 and shipped it to a guest list first. Google says vetted cyber defenders in its Fairwind Program get the model now, while paid API customers and Google AI Ultra subscribers come next with no date attached, so as of October 8 Argon has a published price and no way for an enterprise buyer to pay it. That price doesn't undercut the other two labs. Argon costs $2 per million input tokens and $10 per million output during an introductory period, the same as GPT-6 Sol, and $4 and $20 after it ends, the same as Opus 5.5.

What GTM operators need to do next week

Enterprise AI vendor contracts: If your Q4 renewal is with a single frontier lab, add a repricing clause that lets you renegotiate once Google publishes Gemini 4 pricing. Vendors will push back, but without the clause you may watch a competitor sign the same deal with that protection and get a better price 90 days from now.

AI-native SaaS unit-economics models: Rebuild the 2027 gross-margin projection assuming three labs competing on price. The Harvey open-weight migration math still holds but the break-even point has shifted. With three labs cutting prices, frontier APIs may stay cheaper for longer, so switching to open-weight models pays off later than it did last week.

Antitrust exposure disclosures: If you’re a public company with frontier-lab partnerships, review the coordination risk language in your next quarterly or annual filing. A third lab in the picture could widen the scope of discovery, so get your legal team's view before the filing closes.

What this means for enterprise AI buyers

When Opus 5.5 shipped at 6:06 pm Eastern on Monday, September 22, the pacing narrative survived on one argument: maybe Google would honor the framework even if Anthropic and OpenAI did not. Google's Gemini 4 landing next ends that argument.

It just took 12 days to go from the essay to three competing releases. Twelve days is a short essay-to-triangulation cycle. That means the antitrust case, enterprise AI contracts, and AI-native SaaS financial models all need updating within the same 90 days.

Wednesday’s 10-day slowdown is now a 12-day one, with three labs instead of two. From here on, the labs’ release calendars are a better guide to AI pricing than any pledge about the pace.

Sources

• The Information: Google nears release of flagship Gemini 4 AI Model (2026-09-23)
• Anthropic: Amodei: We Must Pace the Frontier (2026-09-12)
• NYT: Anthropic ships Claude Opus 5.5 (2026-09-22)
• CNBC: Anthropic and OpenAI both released cheaper models (2026-09-22)

From here on, the labs’ release calendars are a better guide to AI pricing than any pledge about the pace.

If your Q4 renewal is with a single frontier lab, add a repricing clause that lets you renegotiate once Google publishes Gemini 4 pricing.

By Business of GTM