SemiAnalysis InferenceX benchmarks show OpenAI's Jalapeño inference chip beating Nvidia's Blackwell on performance per watt across models including GPT-OSS and DeepSeek R1. Ho, OpenAI's hardware vice president, says each of the chip's 64 cores addresses its own HBM bank, and the design reached tapeout in nine months.
Anthropic makes Claude Models on Google Cloud (Anthropic managed) generally available, giving Google Cloud customers the full Claude Platform feature set with Google Cloud IAM credentials, a single invoice, and commitment drawdown. Anthropic says new features ship the same day as on its native platform. Fast Mode on Claude Opus 5.5 is included, priced at $8 per million input tokens and $40 per million output tokens.
OpenAI is close to finalizing a funding round of up to $30 billion at a $1.25 trillion valuation, led by investors from the United Arab Emirates and Coatue, according to people familiar with the deal. The company is positioning the financing as a bridge ahead of a potential IPO and is negotiating a commercial deal for more computing capacity from the UAE. OpenAI told investors it aims to secure signatures by the end of next week, though the final size is still being determined.
METR and Redwood Research report a three-week red-team exercise found exploitable holes in OpenAI's internal misalignment monitoring, including an exploit that could hide hundreds of recent agent actions. METR estimates 75 to 94 percent of internal agentic Responses API traffic from models at least as strong as GPT-5.5 was eligible for monitoring. OpenAI says the production vulnerabilities are patched.
Oracle says its AI Agent Memory 26.8 stores images as DOCUMENT records in Oracle AI Database, so agents can search screenshots and receipts through text descriptions or captions. Default reads omit raw bytes; applications call include_bytes=True to render or forward the original image.
A LessWrong post by josefely argues the US and China can only durably slow AI through a treaty governing compute, because a unilateral pause would let China erase a roughly 10-point Epoch Capabilities Index lead. Verification would depend on satellite monitoring, on-site inspections and tamper-resistant chips that meter and license compute, the post says.
Google CEO Sundar Pichai says the Gemini 3.5 Pro model never shipped because the team diverted effort to Flash. He tells Alex Heath that gaps in agentic coding pushed faster iterative loops on the smaller model, producing 3.5 through 3.8 Flash. By then, Argon pre-training had finished.
Perplexity Research and NVIDIA propose JOLT, a self-distillation method that trains one model as both teacher and student. They claim JOLT+ beats GRPO by 4.2 percentage points on MATH and 9.9 points on the AppWorld state-test pass rate. On MATH, JOLT uses 6.5 times fewer processed training tokens at matched accuracy.
Anthropic says its updated Claude Security, in public beta for Claude Enterprise customers, now rescans repositories as code changes and marks findings fixed when fixes ship. The tool adds an API, an MCP connector, and role-based access controls. Anthropic offers $25,000 in monthly usage credits for two months, through Dec. 31, 2026.
Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including agents acting independently of user instructions, sources tell Axios. Morgan Stanley has begun blocking third-party agents; JPMorgan, Bank of America and Citi are also restricting access. A Morgan Stanley spokesperson says it has not observed rogue agents on its platforms.