Urgent.News

What's breaking now, across thousands of outlets.

AI

Claude Opus 5.5 ถูกลง 40 เปอร์เซ็นต์แต่เก่งขึ้น: สงครามโมเดลยุคใหม่เริ่มแล้ว

โดย Nokka (นก-กา) | 23 กันยายน 2569 บทความนี้เขียนโดย AI (โมเดล glm-5.3 ของผู้ให้บริการ ollama-cloud) ผ่าน Hermes Agent จาก Nous Research ตรวจสอบและเรียบเรียงโดย Nokka Claude Opus 5.5: ถูกลงมาก แต่เก่งขึ้น โมเดลแนวหน้าราคาไม่แนวหน้าแล้ว วันที่ 22 กันยายน บริษัท Anthropic ปล่อยโมเดลระดับแนวหน้าตัวใหม่ Claude Opus 5.5 ออกมา พร้อมข้อความที่ฝืนความเชื่อเดิมของวงการ:…

Anthropic released Claude Opus 5.5, a more cost-effective model, after unveiling Fable 5.1. The new model offers impressive performance at a significantly lower price point, sparking a fierce competition in the model market. In addition to the price reduction, Opus 5.5 demonstrates remarkable efficiency, using fewer tokens to complete tasks compared to its predecessor, resulting in a 40% reduction in overall costs.

This efficiency gain is particularly noticeable in tasks requiring multiple readings of the same context, making it an attractive option for businesses seeking to optimize their AI-driven workflows. However, the model's enhanced performance is not the only noteworthy aspect of its release. Anthropic also highlights significant improvements in security measures, with Opus 5.5 undergoing rigorous third-party testing by Frontier Design and METR, scoring among the highest in a suite of 2,000+ autonomous behavior tests.

These security enhancements, coupled with the model's superior capabilities, position Opus 5.5 as a strong contender in the rapidly evolving AI landscape.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

CLOSEDQUORUM: An Autonomous Windows Implant Using Multiple LLMs for Attack Decisions

+09:00 Original Source : Cisco Talos Severity: high Severity Rationale: Talos statically analyzed a Windows implant that sends host information to up to four commercial LLMs and automatically executes…

  • CLOSEDQUORUM implant sends host info to multiple LLMs for attack decisions
  • Collects hostname, OS, CPU count, admin privileges, target processes
  • Executes credential dumping, browser credential collection, process injection

Real‑Time AI Video Avatars: Your 2026 Guide for Zoom, Teams & Streamers

Real‑Time AI Video Avatars Are Taking Over 2026: A Practical Guide for Zoom, Teams, Discord & Streamers Introduction The moment you click “Join Meeting” in 2026, your webcam will probably be replaced…

  • Real-time AI avatars set to replace webcams on Zoom, Teams, Discord, streaming services
  • Meta and Apple offer ready-made solutions for AI avatars

More from Wednesday 23 September →