|
|
|
Apr 24, 2026
|
|
|
|
|
Supported by
|
|
|
|
|
|
|
Happy Friday! OpenAI releases 'Spud' GPT-5.5 model. Cursor last month hit $2.7 billion in annualized revenue. DeepSeek releases its next-generation V4 model after inital delays.
|
|
|
|
OpenAI on Thursday released its latest flagship model, GPT-5.5, according to a blog post. That model, codenamed Spud internally, is better at breaking down complicated requests from customers and showed improvement in coding, financial modeling and scientific research, compared to prior models from the ChatGPT maker. OpenAI hopes the model will help seize attention and momentum from archrival Anthropic, which has nearly caught up to OpenAI in terms of revenue on the strength of its AI models for coding and is getting global attention for Mythos, an unreleased model with
strong cyber defense and attack capabilities. OpenAI also said its new model is significantly faster in producing responses, thanks in part to its engineers who used OpenAI’s Codex coding tools to automatically find and make improvements to the model’s infrastructure and code, the blog post said. GPT-5.5 also uses significantly fewer tokens to complete the same tasks compared to previous OpenAI models, the company said. That can make the model both faster and cheaper.
|
|
|
|
AI coding startup Cursor, which recently agreed to a potential $60 billion acquisition by SpaceX, last month hit $2.7 billion in annualized revenue, 14 times higher than that revenue pace a year ago, The Information reported Thursday. Some investors expected annualized revenue to hit more than $7 billion by year’s end. Cursor’s revenue should deepen the AI revenue at SpaceX. In its last fiscal year, which ended in January, Cursor generated about $770 million in total revenue, or almost a quarter of the revenue generated last year by SpaceX’s AI division, which
now includes xAI, according to the report. Meanwhile, the startup’s gross margin—revenue after accounting for the core cost of its product—was negative 23% as of the quarter ended in January. It has recently turned positive. And Cursor lost nearly $900 million in its last fiscal year.
|
|
|
|
DeepSeek on Friday launched V4, its highly anticipated new series of open-source AI models with enhanced reasoning and coding capabilities, after inital delays. This is the first new-generation model that DeepSeek released since its R1 model became a global sensation in January 2025. DeepSeek announced two new models, DeepSeek-V4-Pro, a large-size model with 1.6 trillion parameters, and the smaller V4-Flash. It said a version of the V4-Pro model outperformed Anthropic’s Claude Opus 4.6-Max on some benchmarks for reasoning, coding and AI agent capabilities. DeepSeek said its own employees have been using V4 for AI coding. Based on the employees’ feedback, the quality of V4’s delivery is close to that of the non-thinking mode of Claude Opus 4.6, but still behind the thinking mode of Opus 4.6, DeepSeek said. The new V4 models are designed to process extensive text while significantly reducing memory and computing requirements compared to previous generations. The V4 launch comes at a critical time for DeepSeek, as the company, owned by Chinese hedge fund High-Flyer Capital Management, recently started discussions to raise outside capital for the first time. The Information earlier this week reported that Tencent Holdings and Alibaba Group are in talks to invest in DeepSeek at a valuation higher than $20 billion. The company said the price of the V4 Pro model’s application programming interface will drop significantly after clusters of Huawei’s Ascend 950 chips go online later this year. That confirmed The Information’s earlier report that the V4 models would run on Huawei’s chips out-of-box. DeepSeek initially planned to launch its V4 models in February, but the release was delayed
several times due to technical and engineering issues.
|
|
|
|
Meta Platforms has made its plans to cut 10% of its staff, or around 8,000 people, official. In a memo to employees, the company’s chief people officer, Janelle Gale, said the layoffs will happen on May 20, Bloomberg reported. Meta also will not fill about 6,000 open jobs at the company. A Meta spokesperson confirmed the accuracy of the Bloomberg report. Reuters last week reported on the company’s layoff plans. In the memo, Gale said the company was making the cuts in part because of Meta’s investments. While she
didn’t explicitly mention AI, the company has previously forecast that it will spend between $115 billion and $135 billion in capital expenditures this year, a 60% to 88% jump from the prior year, much of which will go to build out its AI computing capacity. “We’re doing this as part of our continued effort to run the company more efficiently and to allow us to offset the other investments we’re making,” she wrote, according to a copy of the memo obtained by Bloomberg.
|
|
|
|
Users of Grok, the AI chatbot operated by SpaceX subsidiary xAI, reported widespread outages on Thursday. Outages were especially common in New York, Illinois and California, according to the tracking site DownDetector. While it’s unclear exactly how widespread the outages are, many Grok users on Reddit and DownDetector were reporting being unable to access the chatbot. For many users, the chatbot displayed a message: “Grok is under heavy usage right now. Please try again later or upgrade your plan to get priority access.” xAI sells premium Grok subscriptions for up to $30 per month. SpaceX has been trying to cut costs at xAI, which operates several large data centers and was acquired by the space firm in February, ahead of the company’s initial public offering later this year. CEO Elon Musk has installed new
leadership at xAI in recent weeks and ordered the team to find ways to cut spending, The Information has reported.
|
|
|
|
| |