From new workhorse models and always-on agents to agent commerce and frontier-AI safety, the period since our last edition brought major developments across the AI-agent stack.
Here’s everything worth catching up on in the latest edition of AI Agents Pulse:
1️⃣ OpenAI released GPT-6.1 Sol for agentic coding, computer use and professional work, making it available in ChatGPT Work, Codex and the API, though not yet in Chat. Standard API pricing is $2 per million input tokens and $10 per million output tokens.
openai.com/index/introducing…
2️⃣ Anthropic released Claude Opus 5.5 and Sonnet 5.5, positioning Opus for complex work and Sonnet for well-scoped everyday tasks. Standard API prices per million input/output tokens are $4/$20 for Opus and $2/$10 for Sonnet.
anthropic.com/claude-opus-5-…
anthropic.com/claude-sonnet-…
3️⃣ SpaceXAI released Grok 4.7 for coding and knowledge work in Cursor, Grok Build and the API. Pricing starts at $2 per million input tokens and $6 per million output tokens.
x.ai/news/grok-4-7
4️⃣ OpenAI introduced Dots, always-on agents powered by GPT-6 Astra that work on their own cloud computers. The rollout began with Pro and Business Premium users in eligible markets, while Enterprise users can try the beta when their workspace administrator enables it.
openai.com/index/introducing…
5️⃣ Microsoft announced Copilot Home, Code and Autopilot, describing Autopilot as a persistent cloud agent with its own identity, memory and workspace. Microsoft planned a phased Frontier rollout for Home and Code and an expanded private preview for Autopilot.
blogs.microsoft.com/blog/202…
6️⃣ Meta announced Enterprise Platform, a new business initiative to bring Muse, Meta Business Agent, Muse API and Muse Code to companies and developers.
about.fb.com/news/2026/09/la…
7️⃣ OpenAI added computer use to the Agents API, letting developers build agents that interact with software to complete tasks. OpenAI and AWS also introduced Bedrock Managed Agents in preview, with both the agent runtime and model inference staying inside AWS.
openai.com/index/devday-2026…
aws.amazon.com/bedrock/manag…
8️⃣ Shopify added WebMCP support for checkout, letting browser agents inspect checkout information, update supported fields and submit orders after buyer confirmation. Agents hand control back when the buyer needs to complete authentication.
shopify.dev/changelog/posts/…
9️⃣ NVIDIA introduced Open Agent Safety Platform, combining OpenShell’s runtime policy enforcement with the Sentry reference design. Sentry uses BlueField-4 DPUs for out-of-band monitoring and containment of agents.
investor.nvidia.com/news/pre…
🔟 OpenAI said it had paused training and evaluation involving tool use for its most capable models pending additional safeguards. It also proposed safety cases covering alignment, containment and monitoring for frontier reinforcement learning.
openai.com/index/how-we-will…
openai.com/index/towards-saf…
1️⃣1️⃣ OpenAI, Anthropic, Google, Meta, xAI and NVIDIA signed the White House Accord on Super Intelligence. The accord sets out a voluntary safety commitment rather than a new mandatory law.
truthsocial.com/@realDonaldT…
1️⃣2️⃣ AMD agreed to acquire World Labs, co-founded by Fei-Fei Li, in an approximately $8.2 billion all-stock transaction. The deal is expected to close by the end of 2026, subject to regulatory approvals and other closing conditions.
ir.amd.com/news-events/press…