AI Daily — November 18, 2026

AI Daily — November 18, 2026
Ai generated image: GPT-6.2 Astra & GPT-6.2 Sol

Models & Research

OpenAI Introduces GPT-6.2 Astra and Sol for More Capable, Cost-Efficient Work — OpenAI introduces GPT-6.2 Astra for complex professional projects and GPT-6.2 Sol for everyday tasks. The company reports improvements in coding, scientific research, document creation, visual design, and multistep workflows, with both models delivering stronger results using fewer output tokens. Astra offers better template matching, cleaner code, and more precise computer use, while Sol balances capability and cost for routine work. Astra’s API input and output prices are 20% lower than its predecessor’s; Sol’s prices remain unchanged. Both models launch in ChatGPT Work, Codex, and the API. OpenAI also describes tighter safeguards for their increased cybersecurity and research capabilities, while acknowledging remaining issues with disclosing blocked tasks. OpenAI ↗

My takeaway: The metric to watch is cost per task. According to Artificial Analysis, GPT-6.2 Sol (max) and GPT-6.2 Astra (max) both cost over 50% less per task than either Claude Opus 5.5 (max) or Claude Fable 5.2 (max) on its intelligence index benchmark. In particular, GPT-6.2 Sol (max) costs $1.01 per task which is about 75% less than GPT-6.2 Astra (max). If you're using GPT-6.1 Sol, the newer model looks like a strong upgrade candidate. Evaluate it on your own use case before switching.

Tools & Open Source

Google broadens access to Gemini Tasks — Gemini’s task mode can call businesses, book appointments, sign into websites, and complete purchases on users’ behalf. Google is expanding access, with some advanced features initially limited to the U.S. TechCrunch AI ↗

My takeaway: Authentication, bot policies, and fraud controls should account for legitimate agents acting on users' behalf.

Nvidia announces Nemotron for Science and Rubin HPC GPU at SC26 — At the SC26 supercomputing conference, Nvidia’s Ian Buck announced tools for AI agents that can plan experiments, run code, and verify results on researchers’ own systems, alongside the Rubin HPC GPU and an expanded Vera CPU platform. Nvidia Blog ↗

My takeaway: Plan on-premises agent infrastructure around both GPU inference and the CPU capacity needed to run tools and evaluate results.

Industry & Funding

  • Nvidia in turmoil as Jensen Huang exits; company says it won't reappoint him
  • Nvidia drops China sales and reaffirms open-source plans
  • Isomorphic Labs raises $5B Series C
  • Amazon and Google to hold nearly 30% of Anthropic after IPO
  • Model ML raises $250M for AI agents in investment banking
  • Humyn Labs raises $47M to supply robotics training data

Summaries are AI-generated and may contain errors — always verify against the linked original. Each story links to its source, which holds the copyright. Outlet names are shown for attribution only and do not imply any endorsement or affiliation.

Disclaimer: The views expressed in My Takeaway are my own personal opinions and general observations on industry trends. They are not intended to criticize, disparage, or make factual claims about any specific company, product, or platform. Any platform names mentioned are referenced solely for illustrative and informational purposes.