Anthropic starts its IPO roadshow

Anthropic began its roadshow on 10 November. It's offering 825 million Class A shares at an expected $70 to $78 each. Of those, 801.35 million are new shares from the company and 23.65 million come from existing holders. Underwriters also get a 30-day option on another 123.75 million shares. It has applied to list on the Nasdaq Global Select Market under the ticker ANT, and Goldman Sachs and Morgan Stanley are handling prospectus requests. The S-1 was filed with the SEC on 26 October. RTTNews adds a planned $26 billion private placement of a separate share class at the IPO price, with AMD, Broadcom, Google, Micron, Microsoft, NVIDIA and TSMC Partners among the buyers.

For builders: A listed Anthropic will publish its numbers every quarter, so teams relying on Claude will get a much clearer view of the company's finances and spending plans.

Claude Fable 5.2 and Mythos 5.2 ship

On 9 November Anthropic released Fable 5.2 and Mythos 5.2. Both are the same underlying model. Fable has tighter cyber and biology safeguards and is open to everyone. Mythos is limited to vetted organisations in Anthropic's verification programs. Fable 5.2 scores 69.9% on FrontierSWE v2, up from 56.3% for Fable 5.1, and 41.3% on the Remote Labor Index, up from 17.9%. Standard pricing is $8 per million input tokens and $40 per million output tokens. An optional fast mode doubles both prices and generates tokens up to 2.5 times faster. Anthropic says the model triggers safety classifier blocks about 55% less often than 5.1.

For builders: The API ID is claude-fable-5-2, and the model is live on AWS, Google Cloud and Azure. Anthropic says the High effort setting matches Opus 5.5 at Max on FrontierCode for about half the cost per task, so it's worth re-testing your effort settings.

Google releases Gemini 4 Hydrogen

Google DeepMind announced Gemini 4 Hydrogen on 12 November. It's a fast model for routine coding and agent work, and it sits below the larger Carbon and Argon models. Google says it finishes tasks 2.5 times faster than Gemini 3.8 Flash at the default high thinking level and makes fabricated claims more than five times less often. It scores 57% on Terminal-Bench 4.1, against 18.2% for 3.8 Flash. Pricing is a flat $1.50 per million input tokens and $7.50 per million output tokens across the full 1 million token context window. A new xhigh reasoning tier joins low, medium and high.

For builders: Long-context jobs get simpler to cost because the price stays the same at any prompt length. The model is available now through the Gemini API, AI Studio and Android Studio.

OpenAI closes $11 billion at a $1.25 trillion valuation

Bloomberg reported that OpenAI received about $11 billion by 6 November. That's the first close of a round valuing the company at $1.25 trillion. Coatue and T. Rowe Price led this close, and the full round could reach $20 billion. Earlier, OpenAI had aimed for at least $30 billion at around $1.4 trillion. The new figure is still about 47% above the $852 billion valuation from its March round.

For builders: OpenAI's lower valuation and Anthropic's public listing are both signs of how much investors want to pay for frontier labs. Pricing and capacity at those labs depend on that money.

Nvidia's board opens a review of Jensen Huang

On 8 November Nvidia said a special committee of independent directors, working with outside counsel, would look into whether its CEO had been targeted by a foreign intelligence operation. The board says it has confidence in Huang and that he'll keep running the company. Shares fell 4.5% in US premarket trading on 9 November. Nvidia reports third-quarter results on 17 November.

For builders: For now this is a governance story, and Bernstein expects little direct effect on demand. Teams planning GPU purchases should still watch the 17 November results and any congressional follow-up.

New York opens RAISE Act registration

New York's new DIGIT office, part of the Department of Financial Services, opened its RAISE registration portal on 10 November. Frontier developers with more than $500 million in annual revenue, which train models with over 10^26 operations, have to register by 1 January 2027. From that date they'll have to report critical safety incidents within 72 hours, publish and maintain a safety framework, and release a transparency report with each new model.

For builders: Most teams won't meet the threshold. Still, the vendors they rely on will start publishing safety frameworks and transparency reports, and those are useful inputs for procurement and risk reviews.

OpenAI and Anthropic put agents in front of customers and meetings

On 10 November OpenAI opened a beta of Businesses in ChatGPT. It includes a Business Hub for verified company profiles and product feeds, and Business Agents that answer customer questions inside ChatGPT. For now it's limited to eligible US e-commerce and online service businesses. On the same day Anthropic announced Claude Meetings, a Team and Enterprise beta that joins Google Meet calls as a visible attendee. It also launched Whiteboards and Sheets betas for paid plans.

For builders: ChatGPT now works as a customer-facing channel, so product data feeds matter more. Claude Meetings leaves a call when an external guest joins unless an admin approves it, which should make it easier to get past internal security reviews.

Sources

Talk about it with people building with AI

Our Auckland meetups are open to everyone. Members get a ForgeBox and the members' projects area for NZ$599 a year.

Become a member

Get the daily AI news by RSS