OpenAI’s Dots is an always-on AI agent that keeps working after you close the laptop — and it’s racing Meta’s Muse in a new privacy battle. Here’s what Dots actually does, what it costs, and whether it’s worth your attention.
Meta unveiled Muse Wonder and Muse Wonder Full on October 12, 2026 — its first models it claims are truly frontier-class, built for agents that can handle real-world tasks.
Continuum AI has launched OrcaCyber Zero 1.5, a gated cybersecurity reasoning model with a 1M-token context window that claims 100% on Cybench — plus plans for an open-weight release next week.
Cisco is putting AI agents from both Anthropic and OpenAI inside Webex — so the agents come to your meeting instead of making you go to them. Plus RoomOS 27 with Google Meet support in November and a customer-experience harness called Dialog.
Odyssey has opened a public research preview of its Odyssey-3 world model, which generates interactive environments in real time — and could let robots, drones, and game agents share one brain.
Microsoft has introduced Decision-1, a compact decision-making model built on Qwen3.5-9B that it says is roughly 35 times faster than GPT-6 Sol on routine AI-agent decisions — priced at $0.042 per million input tokens, with output tokens free. Here is what the launch means, and what Microsoft still needs to prove.
GitHub’s Copilot Studio, launched in late September, chains Copilot agents into event-triggered visual workflows — no Actions YAML required.
Kling 4.0 entered early access on September 30 with native 30-second clips, 10 keyframe images, 15 reference assets, and two-channel stereo sound.
OpenAI’s GPT Image 2.5 splits into Flare and Sunburst tiers — faster by default, slower and more precise when it counts. Same token pricing.
DeepSeek’s V4.1 Flash pairs a 552B-parameter MoE with an MIT license and 1M-token context. What the open-weights efficiency release changes.
