Claude Opus 5.5 · Terminal-Bench-Science(0.1) 58.7% · 主文证据(ev-op55-terminal-science)
Claude Opus 5.5 的 Terminal-Bench-Science · 0.1 官方出处。Adaptive Max · 58.7%。见《Introducing Claude Opus 5.5》本地留存副本 lines 37-38, 46; official HTML table。证据 ev-op55-terminal-science。
Claude Opus 5.5 · Terminal-Bench-Science(0.1),Adaptive Max配置,Score 58.7%(具体成绩公布时间未确认)。见《Introducing Claude Opus 5.5》本地留存副本 lines 37-38, 46; official HTML table。 查看成绩条件
Terminal-Bench-Science 0.1 — Opus 5.5: 58.7%; standard error ±3.5–5 points per model.
Claude Opus 5.5 · Terminal-Bench-Science(0.1),Adaptive Max配置,Score 58.7%(具体成绩公布时间未确认)。官方出处 Introducing Claude Opus 5.5(主文,官方发布页;本地保存 HTML/TXT:本地留存副本、本地留存副本;主表为 HTML;原文发布日期 2026-09-22,当前修订时间未知)本地留存副本 lines 37-38, 46; official HTML table,https://www.anthropic.com/news/claude-opus-5-5#performance-and-cost-effectiveness。Benchlog https://benchlog.lekshome.cc/evidence/claude-opus-5-5/ev-op55-terminal-science/
System Card: safeguards enabled; flagged requests use default server-side fallback (3.9% of requests, affecting 5% of trials); ten trials per task, 700 total. Standard error is ±3.5–4.8 points; do not merge with other harnesses.
Claude Opus 5.5 · Terminal-Bench-Science (0.1) 58.7% · Main text evidence (ev-op55-terminal-science)
Official source for Terminal-Bench-Science · 0.1 in Claude Opus 5.5. Adaptive Max · 58.7%. See “Introducing Claude Opus 5.5” at 本地留存副本 lines 37-38, 46; official HTML table. Evidence ev-op55-terminal-science.
Claude Opus 5.5 · Terminal-Bench-Science (0.1), configuration Adaptive Max, Score 58.7% (exact score publication date unconfirmed). See “Introducing Claude Opus 5.5” at 本地留存副本 lines 37-38, 46; official HTML table. View score conditions
Terminal-Bench-Science 0.1 — Opus 5.5: 58.7%; standard error ±3.5–5 points per model.
Unconfirmed · Launch or later disclosure unconfirmed
Collected
Historical basis
No reliable historical version evidence available
Citation text
Claude Opus 5.5 · Terminal-Bench-Science (0.1), configuration Adaptive Max, Score 58.7% (exact score publication date unconfirmed). Official source: Introducing Claude Opus 5.5 (Main text, 官方发布页;本地保存 HTML/TXT:本地留存副本、本地留存副本;主表为 HTML;原文发布日期 2026-09-22,当前修订时间未知), 本地留存副本 lines 37-38, 46; official HTML table, https://www.anthropic.com/news/claude-opus-5-5#performance-and-cost-effectiveness. Benchlog https://benchlog.lekshome.cc/evidence/claude-opus-5-5/ev-op55-terminal-science/?lang=en
Related scores and conditions
Claude Opus 5.5 · Terminal-Bench-Science scores and configurations
System Card: safeguards enabled; flagged requests use default server-side fallback (3.9% of requests, affecting 5% of trials); ten trials per task, 700 total. Standard error is ±3.5–4.8 points; do not merge with other harnesses.