AI & Machine Learning
350 articles RSS
OpenAI's GPT-6.2 Astra and Sol Beat Benchmarks and Cross a Self-Improvement Threshold, but Power Users Call Astra Unusable
OpenAI's GPT-6.2 Astra tops independent benchmarks and crosses a new AI self-improvement threshold, but early power users call it unusable for long agentic work.
GitHub Copilot's Ignite-Week Release Pairs a Self-Reported Terminal-Bench Lead With Browser Control and Enterprise Agent Monitoring
On November 17, GitHub shipped new coding models, a vendor-claimed Terminal-Bench lead, browser control in preview, and Agent 365 audit monitoring in GA, the same week Microsoft's 365 Copilot debuted its own autonomy-plus-audit feature.
A 2024 Hallucination Bound Now Anchors Three 2026 Papers on LLM Training Data, One Extending It to Scientific Discovery
A 2024 theorem tying LLM hallucination to once-seen training facts has drawn three 2026 papers, one linking it to scientific discovery and another challenging standard deduplication practice.
Arena, Artificial Analysis and Terminal-Bench Disagree on Which AI Model Leads This Month
Arena, Artificial Analysis and Terminal-Bench each credit a different model at the top, and researchers say the disagreement traces to what each benchmark actually measures.
Fal Raises $1.3 Billion Series F at $16 Billion Valuation as AI Media-Generation Demand Outpaces Its Compute Supply
The generative-media inference startup doubled its valuation to $16 billion after demand for its image, video and audio models forced it to rate-limit customers.
Perplexity Releases Decider v1.3 and Flash v1.3 Weights Under Apache 2.0; Artificial Analysis Decision Index Lists v1.3 Above Jev 1.14, Flash Below
MarkTechPost reports Perplexity's open-weight Decider v1.3 and Flash v1.3 ship under Apache 2.0 on Qwen bases; Artificial Analysis's Decision Index scores v1.3 at 70.41, Jev 1.14 at 66.23 and Flash v1.3 at 64.07.
Self-hosted Claude Managed Agents is in private beta; its license gives licensees 8 hours to disable a feature or roll back after a safety notice
Anthropic's license for the self-hosted Managed Agents server sets 8 hours to disable a feature or roll back after a written risk notice. The New Stack reports it; Anthropic's docs list what the server sends.
Cursor Adds Multiplayer to Projects in Beta on November 10; Docs Say Teammates' Coordinators Message Each Other and Treat Messages as Untrusted Input
Cursor says Projects is now multiplayer, in beta. Its docs require a team account and Cursor 3.26.13 and say a receiving agent treats a teammate's agent message as untrusted input; The New Stack contrasts Google's /swarm preview.
Google Rolls Out Orchestrator in Antigravity, Pairing a Carbon or Argon Planner With Hydrogen; Two Google Posts Name Different Terminal-Bench Versions
Google says Orchestrator, rolling out in Antigravity on November 12, hands a task between a Gemini 4 planner and Hydrogen; Google's pages differ on the Terminal-Bench version, the planner options and enterprise access.
MCP Working Group Publishes an Experimental Draft for Moving Files Between Clients and Tools Outside Model Context; No Reference Implementation Yet
An experimental MCP extension draft, io.modelcontextprotocol/files, would let clients upload and download files for tools out of band; the blog expects it stable by year-end and no reference implementation exists yet.
Artificial Analysis Scores Google's Gemini 4 Hydrogen 51 Against Argon's 50 at High Thinking, With $1.87 per Index Task Versus $4.97
Google's Gemini 4 Hydrogen costs $1.50 and $7.50 per million tokens. Artificial Analysis scores it 51 at high thinking against Argon's 50; Google Cloud lists 3.8 Flash at half that price through December 31.
Google Announces XLA Serving Engine, a C++ Runtime for Exported Models; Its Benchmark Reports 2x to 5x a Python Server's Throughput on Small Requests
Google's blog launches XLA Serving Engine, a C++ runtime for exported models. Its repo benchmark, run on one TPU chip, lists 2.0x to 5.0x a Python server's throughput for 1- and 16-item requests and 1.14x at 128.