Video by Open Data Science and AI Conference via YouTube
Three US organisations released open-weight models on October 15 and 16, with different license terms, hardware needs, and capability.
The most accessible is NVIDIA’s 74B-parameter model, which the company says fits in 24GB of memory. Meta’s Muse Wonder and Google’s Gemini 4 Argon added frontier options with different access limits and prices, and Claude Code Review reached general availability. OpenAI’s new misalignment reports show an agent leaving an action out of its own summary. Jobs covers Anthropic’s data on junior hiring, Oyster’s count of postings that require AI skills, and surveys from KPMG and Selby Jennings. The episode also covers Claude-designed quantum circuits, Anthropic’s draft federal safety bill and evaluator access for METR, and Accenture moving extra Claude costs onto team budgets.
In this episode
NVIDIA releases Nemotron 3.5 Bolt, a 74B model that NVIDIA says fits in 24GB, alongside Nemotron 3.5 Super VL.
Meta releases Muse Spark Open, an 816B-parameter model, under a custom license.
Reflection AI releases Beam weights under Apache 2.0.
Meta introduces Muse Wonder, with its higher-scoring Wonder Full model held back from the API.
Google opens Gemini 4 Argon to developers, enterprises and subscribers, with introductory pricing through November 30.
Claude Code Review reaches general availability for Team and Enterprise plans.
OpenAI adds four reports to its misalignment tracker, including the Sepolia wallet case.
Anthropic reports a drop in junior technical hiring from 27% to 9%.
Oyster finds postings requiring AI skills rose from about 5% to 15% between 2022 and the first half of 2026.
KPMG and Selby Jennings surveys tie AI skills to hiring and pay.
Claude designs quantum circuits for eight tasks with Harvard and QuEra collaborators.
Anthropic pitches a draft federal frontier AI safety bill to lawmakers.
Anthropic gives METR employee-level access and outlines a three-tier evaluator structure.
Accenture moves extra Claude costs to team budgets.
Quick hits: OpenAI’s Rosalind Academic Plan, Vals AI’s impossible refund test, XBOW’s Chrome bounty, the Tencent Cloud agent fleet probing Amap, Modal’s Series D, Anthropic’s Nova Scotia lease, Schwab and JPMorgan blocking outside agents, the disputed Replit and Anthropic relationship, Character.AI’s Chapter 11 filing, and Brad Smith on immigration and AI talent.
ODSC updates: ODSC AI West 2026 and ODSC AI NYC, the AI for Work Summit.
This episode was sponsored by:
ODSC AI West 2026 returns October 27 to 29 at the Hyatt Regency San Francisco Airport in Burlingame, with more than 250 speakers across hands-on training in machine learning, LLMs, agentic AI, and AI engineering. Learn more: odsc.ai/west
The ODSC AI Engineering Accelerator is a structured programme for engineers building production AI systems. Learn more: https://odsc.ai/accelerator/
ODSC AI NYC, the AI for Work Summit, runs December 2 to 3 at the Jay Conference Center at Bryant Park, focused on applied enterprise AI: agents and copilots, deployment, evaluation, governance, and production workflows. Learn more: summit.ai
To watch more videos like this, visit https://aiplus.training
Sign up for the newsletter to stay up to date with the latest trends in data science: https://opendatascience.com/newsletter/
Follow us online!
• Facebook: https://www.facebook.com/OPENDATASCI
• Instagram: https://www.instagram.com/odsc/
• Blog: https://opendatascience.com/
• LinkedIn: https://www.linkedin.com/company/open-data-science/
• X (twitter): https://x.com/_odsc