Back to Blog

AI News Roundup – 6 November 2026: Anthropic IPO Buzz, Perplexity Agents, OpenAI Benchmarks & More

AI News
AI News Roundup – 6 November 2026: Anthropic IPO Buzz, Perplexity Agents, OpenAI Benchmarks & More

Today's AI news: Anthropic IPO valuation hits $1.1T, Perplexity launches work agents, OpenAI's BioBasicsBench, and Inkling 2 Small's training run.

••4 min read

Here is what happened in AI today, Friday 6 November 2026 — a busy day spanning model releases, product launches, a landmark IPO story, and new research benchmarks.

Anthropic IPO Buyers Eye Valuation as Low as $1.1 Trillion

The biggest business story of the day comes from Bloomberg, which reports that investors eyeing Anthropic's upcoming IPO are pricing the company at as low as $1.1 trillion — a figure that reflects pressure on AI valuations after a broader market rout. Anthropic has been one of the most closely watched private AI companies, backed by billions in investment from Amazon and Google among others. A valuation in the low trillions would still make it one of the most valuable technology listings in recent memory, but the downward pressure signals that public market appetite for AI is being tested after a period of frothy private valuations.

Perplexity Launches AI Agents for Work

Perplexity has introduced what it calls Teammates — AI agents designed to handle complex, ongoing work tasks. The launch marks Perplexity's move beyond its search-first identity into the enterprise productivity space, where it now competes more directly with offerings from OpenAI and Anthropic. The agents are built to manage multi-step workflows rather than answer one-off queries, positioning Perplexity as a broader AI work platform. The announcement landed on Hacker News with notable community interest.

OpenAI Launches BioBasicsBench to Measure Wet-Lab AI

OpenAI has introduced BioBasicsBench, a 100-task benchmark designed to measure how well AI agents interpret raw experimental data and handle the everyday tasks of wet-lab science. The benchmark is a direct response to growing interest in AI's role in biological research — agents are tested on the kind of routine but critical data-interpretation work that takes up a large portion of a working scientist's day. This release sits alongside a broader trend of AI labs developing domain-specific evaluation tools to move beyond general reasoning benchmarks.

Inkling 2 Small — 17A-594B Model Trained in 68 Hours on 12 Trillion Tokens

AINews reports on Inkling 2 Small, a 17-active-parameter, 594-billion-total-parameter mixture-of-experts model that completed its training run in just 68 hours on 12 trillion tokens, reportedly beating the competing Muse Spark model on key benchmarks. The same edition covers METR's study of AI research workflows at Anthropic. The training efficiency numbers are striking — completing a frontier-scale run in under three days points to continued rapid progress in infrastructure and training optimisation. Inkling 2 Small's results will likely draw attention from researchers comparing open and semi-open model ecosystems.

Microsoft Winds Down Oracle Cloud Capacity Deal

Reuters reports that Microsoft is winding down an arrangement under which it rented AI cloud capacity from Oracle. The move is a notable shift in the dynamics of AI infrastructure partnerships — Microsoft has been aggressively building out its own Azure AI capacity, and the decision to step back from renting compute from a rival signals confidence in its own data centre expansion. For Oracle, which has been positioning its cloud infrastructure as a key AI compute provider, losing a Microsoft contract is a meaningful development to watch.

OpenAI Partners with Hacktron for Security Testing

OpenAI has announced a partnership with Hacktron — a bug bounty and security research platform — to find vulnerabilities in OpenAI's systems, test connections with third-party services, and strengthen its overall defences using frontier AI. The partnership reflects the growing importance of red-teaming and external security validation as AI systems become more deeply integrated into critical workflows. OpenAI framed the collaboration as part of its ongoing commitment to responsible deployment.

That is today's AI news digest. Check back tomorrow for the next roundup from ItsTimeAI.

Frequently Asked Questions

What is Anthropic's expected IPO valuation in November 2026?

According to Bloomberg, investors eyeing Anthropic's IPO are pricing the company at as low as $1.1 trillion, reflecting pressure on AI valuations after a broader market downturn.

What are Perplexity's new Teammates AI agents?

Perplexity launched Teammates, AI agents designed to handle complex, ongoing work tasks and multi-step workflows, moving the company beyond its search-first identity into the enterprise productivity space.

What is OpenAI's BioBasicsBench?

BioBasicsBench is a 100-task benchmark introduced by OpenAI to measure how well AI agents interpret raw experimental data and handle everyday wet-lab science tasks.

What is Inkling 2 Small and why is it significant?

Inkling 2 Small is a mixture-of-experts model with 17 active parameters and 594 billion total parameters, trained in just 68 hours on 12 trillion tokens. It reportedly beats the Muse Spark model on benchmarks, highlighting rapid advances in training efficiency.

Explore Topics

#ai-news#roundup