Anthropic Lets Claude Orchestrate 1,000 Agents at Once · 1010-728

2026-10-10 16:00

Author: RadarAI Editorial Editor: RadarAI Editorial Last updated: 2026-10-10 Review status: Editorial review pending Brief 速报官方 AI动态开源

The AI agent race is moving from concept to large-scale deployment: Anthropic has launched dynamic workflows for Claude, capable of orchestrating 1,000 agents in parallel in a single run [14]; Meta, together with Shopify, Stripe, and others, has introduced the Personal Agent Protocol (PAP), attempting to standardize how agents interact with commercial services [2]. Meanwhile, OpenAI's unreleased model gpt-rosalind-discovery has appeared on its API pricing page, hinting at a possible deeper push into vertical domains such as drug discovery...

Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.

## 🔍 Core Insights **AI agents** are moving from concept to large-scale deployment: **Anthropic** has launched dynamic workflows for Claude, capable of orchestrating **1,000 agents** in parallel in a single run [14]; **Meta**, together with **Shopify**, **Stripe**, and others, has introduced the Personal Agent Protocol (**PAP**), attempting to standardize how agents interact with commercial services [2]. Meanwhile, **OpenAI**'s unreleased model **gpt-rosalind-discovery** has appeared on its API pricing page, hinting at a possible deeper push into vertical domains such as drug discovery [6]. ## 🚀 Key Updates - **Anthropic launches dynamic workflows for Claude, orchestrating up to 1,000 agents in parallel per run** [14]: A main agent plans and dispatches subtasks, suited for scenarios such as large-scale code review. - **Meta, together with Sierra, Shopify, and others, introduces the Personal Agent Protocol (PAP)** [2]: Standardizes how personal agents interact with enterprise services; Walmart, Stripe, and others have already joined. - **OpenAI's unreleased model gpt-rosalind-discovery appears on its API pricing page** [6]: Priced the same as gpt-rosalind-research, possibly dedicated to drug discovery. - **Cloudflare launches open-weight decision model Clef-omni** [5]: Adds audio and video input, supports processing multimodal data in a single API call, with weights now open on Hugging Face. - **Microsoft releases decision scoring model Microsoft-Decision-1** [9]: Post-trained on Qwen3.5-9B, outputs calibrated probabilities instead of generated text, now available on Microsoft Foundry and OpenRouter. - **Alibaba's Qoder launches Fast mode** [12]: First response time cut from 8 seconds to 3 seconds, Credits consumption reduced to 0.8x, with output quality unchanged. - **Personal ChatGPT subscriptions can now be used directly with Devin** [23]: After connecting a Go, Plus, or Pro subscription to Devin, GPT model usage is deducted from the subscription quota. - **A GEO service provider in Shenzhen fined 50,000 yuan for implanting ads into AI** [1]: The company fabricated fake industry ratings to increase the likelihood of its content being cited by AI, violating anti-unfair competition regulations. ## 🔗 Sources [1] A GEO service provider in Shenzhen fined 50,000 yuan for implanting ads into AI, constituting false advertising — https://aihot.news/items/rm7rulwudkub3gktusz9a0yig [2] Meta, together with Sierra, Shopify, and others, introduces the Personal Agent Protocol (PAP) — https://aihot.news/items/k35hntkgim2vm4bz9qnkjhevd [5] Cloudflare launches open-weight decision model Clef-omni, adding audio and video input — https://aihot.news/items/uyj1nct0n4u1mw332xg97xvay [6] OpenAI's unreleased model gpt-rosalind-discovery appears on its API pricing page — https://aihot.news/items/mnaysrpr4ic61nfshulaqee5v [9] Microsoft releases Microsoft-Decision-1: a decision scoring model based on Qwen3.5-9B — https://aihot.news/items/x1upix3ug5q2zpo4ojrzcfjcb [12] Alibaba's Qoder launches Fast mode: first response in 3 seconds, Credits consumption reduced to 0.8x — https://aihot.news/items/f3d8q8cnr1jy467442zj1fshp [14]

AI agents are moving from concept to large-scale deployment: Anthropic has launched dynamic workflows for Claude, capable of orchestrating 1,000 agents in parallel in a single run [14]; Meta, together with Shopify, Stripe, and others, has introduced the Personal Agent Protocol (PAP), attempting to standardize how agents interact with commercial services [2]. Meanwhile, OpenAI's unreleased model gpt-rosalind-discovery has appeared on its API pricing page, hinting at a possible deeper push into vertical domains such as drug discovery [6].

🚀 Key Updates

Anthropic launches dynamic workflows for Claude, orchestrating up to 1,000 agents in parallel per run [14]: A main agent plans and dispatches subtasks, suited for scenarios such as large-scale code review.
Meta, together with Sierra, Shopify, and others, introduces the Personal Agent Protocol (PAP) [2]: Standardizes how personal agents interact with enterprise services; Walmart, Stripe, and others have already joined.
OpenAI's unreleased model gpt-rosalind-discovery appears on its API pricing page [6]: Priced the same as gpt-rosalind-research, possibly dedicated to drug discovery.
Cloudflare launches open-weight decision model Clef-omni [5]: Adds audio and video input, supports processing multimodal data in a single API call, with weights now open on Hugging Face.
Microsoft releases decision scoring model Microsoft-Decision-1 [9]: Post-trained on Qwen3.5-9B, outputs calibrated probabilities instead of generated text, now available on Microsoft Foundry and OpenRouter.
Alibaba's Qoder launches Fast mode [12]: First response time cut from 8 seconds to 3 seconds, Credits consumption reduced to 0.8x, with output quality unchanged.
Personal ChatGPT subscriptions can now be used directly with Devin [23]: After connecting a Go, Plus, or Pro subscription to Devin, GPT model usage is deducted from the subscription quota.
A GEO service provider in Shenzhen fined 50,000 yuan for implanting ads into AI [1]: The company fabricated fake industry ratings to increase the likelihood of its content being cited by AI, violating anti-unfair competition regulations.

🔗 Sources

[1] A GEO service provider in Shenzhen fined 50,000 yuan for implanting ads into AI, constituting false advertising — https://aihot.news/items/rm7rulwudkub3gktusz9a0yig [2] Meta, together with Sierra, Shopify, and others, introduces the Personal Agent Protocol (PAP) — https://aihot.news/items/k35hntkgim2vm4bz9qnkjhevd [5] Cloudflare launches open-weight decision model Clef-omni, adding audio and video input — https://aihot.news/items/uyj1nct0n4u1mw332xg97xvay [6] OpenAI's unreleased model gpt-rosalind-discovery appears on its API pricing page — https://aihot.news/items/mnaysrpr4ic61nfshulaqee5v [9] Microsoft releases Microsoft-Decision-1: a decision scoring model based on Qwen3.5-9B — https://aihot.news/items/x1upix3ug5q2zpo4ojrzcfjcb [12] Alibaba's Qoder launches Fast mode: first response in 3 seconds, Credits consumption reduced to 0.8x — https://aihot.news/items/f3d8q8cnr1jy467442zj1fshp [14]

← Back to Updates