Only two models could finish on keys alone.
Claude fable 5.2 → 66/66 in 3:07:50 claude opus 5.5 → 66/66 in 5:13:44 gpt-6 astra got close but it ran out of time at 64. sonnet 5.5 made it to 63.
AI 游戏开发日报 · 11.10
今天 5 条： 1. Claude Fable 5.2 发布：$8 / $40 每百万 token，官方 FrontierSWE v2 69.9%（Opus 5.5 62.3%）；Claude Code 2.1.321 当天设为默认 Fable 模型 2. 游戏开发者实测同一需求（设计一个 FPS 新敌人）：Fable 5.2 26 分钟 $10.81，Opus 5.5 15 分钟 $3.65——约 3
Anthropic has officially released Claude Fable 5.2 and Claude Mythos 5.2
Its new frontier models focused on deep technical reasoning, scientific tasks, and autonomous software engineering. Announcement reduces API prices by 20%, expands window to 1 million
Fable 5.2 and other four models tested on the same question Fable
Fable 5.2 and other four models tested on the same question Fable: less than 6 minutes, weekly quota 6% Opus: about 5 minutes Sol: 1 minute 53 seconds Astra: 2 minutes 01 seconds Fable's fire is the word burning, and fly is the word spelled out by a thousand birds; Opus is the same style, with less details. Sol is the most elegant, burned into piles of fallen leaves by fire; Astra is left with only particles.
You have reached the end of the archive
All of Claude Opus 5