FRAMEWIREIndonesiaUpdated Sep 22Live wire
0:00 / 0:00

I gave claude a brain that decides how hard to think

Not per session. Per step. Jev is a classifier that scores each move 0 to 1. Claude Opus 5 shifts between LOW, MEDIUM, and HIGH reasoning mid-run. Reading files. LOW. Twelve steps in a row. Then it cross-references modules.

BuboSep 225
All of Claude Opus 5Archive
0:00 / 0:00

Anthropic拥有的Claude刚刚发布了这个： “介绍Claude Opus 5.5，这是我们新Claude 5.5家族的第一款型号。

对于大多数任务，它的性能都达到Claude Fable 5.1的水平，运行成本比Opus 5低40%。”

美股雷达Sep 22
All of Claude Opus 5Archive
0:00 / 0:00

ChatGPT users watching Opus 5.5 drop

It beats Astra in agentic coding • 1846 vs 1542 on GDPval-AA • ~20% of Astra’s cost on FrontierCode • 30%+ faster than Opus 5 • higher Claude usage limits Anthropic didn’t drop a model. They dropped a bomb on ChatGPT users 😭

PriySep 221
All of Claude Opus 5Archive
0:00 / 0:00

【速報】Claude Opus 5.5をAnthropicが公開、Fable 5.1並みの性能でOpus 5より4割安く

・入力100万トークン4ドル、出力20ドル。Opus 5から2割値下げ ・Sonnet 5.5とHaiku 5.5も数週間以内に公開予定 ・生物・サイバー分野はFable 5.1と同じ利用制限 来年3月末までにClaude 6が公開される確率は34%

Polymarket JapanSep 222
All of Claude Opus 5Archive