AI news · Thursday, August 13, 2026

OpenAI targets real-time business tasks with new 14x faster mode

OpenAI is leaning into enterprise speed with a new preview mode called Ultrafast, which lets GPT-5.6 Sol churn out text at 14 times the standard speed. By hitting 750 tokens per second, OpenAI is trying to solve a perennial trade-off where companies previously had to choose between a 'smart' model or a 'fast' one. This hardware-heavy push, powered by a partnership with chipmaker Cerebras, suggests the company is racing to capture time-sensitive roles like financial analysis and incident response. The announcement lands as the company navigates a major leadership shift: Chief Revenue Officer Denise Dresser is out after nine months, replaced by Dali Rajic, a former executive from Wiz, the security firm Google just picked up for $32 billion. OpenAI’s product ecosystem has grown to over one billion weekly users and two million businesses, but internal pressure remains high to prove that scale translates into sustainable revenue before a potential IPO.

Infrastructure providers are seeing a similar, if arguably more chaotic, surge in activity. A massive $500 billion plan involving Wall Street heavyweights like BlackRock and Goldman Sachs is underway to build out AI data centers, with Nvidia effectively acting as the guarantor for the aging GPUs used as collateral in these deals. The model, which some critics compare to risky dot-com era financing, is designed to ensure there is enough physical compute power to satisfy the explosive demand from hyperscalers. This infrastructure buildout is already showing results, with data center firm CoreWeave reporting 112% year-over-year revenue growth and Nebius surging 450%, signaling that demand for underlying compute is still far outstripping supply.

Microsoft is responding to the maturity of the market by cleaning house. The company is consolidating its consumer and business Copilot apps into a single 'super app' and retiring underperforming features like AI-generated podcasts and its emotive yellow avatar, Mico. The consolidation acknowledges that Microsoft’s original strategy was needlessly complex, making it difficult for the company to compete against the unified experiences of ChatGPT or Gemini. Meanwhile, for those wondering about the future of AI in the workplace, Anthropic is pushing further into the browser by bringing its 'Claude Cowork' agent directly into Chrome. The tool can now handle research, organize files, and create reports within the side panel, though Anthropic still warns that—like all these agents—it remains a potential target for hackers looking to sneak in malicious instructions.

The quick hits

OpenAI launched Ultrafast mode to make its most powerful model perform at 14x its normal speed — it’s a move to capture enterprise workflows that demand instant results.
Nvidia is backing a $500 billion infrastructure plan that guarantees the value of used chips — it aims to keep the AI gold rush fueled by preventing hardware from going obsolete.
Microsoft is retiring its Copilot avatar Mico and merging its separate AI apps — the company is simplifying its tools after admitting the original strategy was too confusing for users.
Anthropic is embedding its Claude agent directly into Chrome side panels — it lets users automate reports and research without ever leaving their browser.

OpenAI targets real-time business tasks with new 14x faster mode

The quick hits

Sources