Anthropic 发布了 Claude Opus 4.8，这是其旗舰 AI 模型的更新版本，特别强调“诚实”和改进的校准。据报道，这一迭代产生静默代码缺陷的可能性降低了四倍，并提供了更明确的不确定性标记，这对于使用自主编码代理的开发人员来说是一个显著的进步。此次发布还包括更快、更便宜的“快速模式”，并在 SWE-Bench Pro 等基准测试中取得进展，表明 Anthropic 的前沿模型正在持续快速的开发周期中。 AI
影响 通过提高可靠性和减少静默错误，加速了自主编码代理的采用。
排序理由 来自前沿实验室（Anthropic）的新模型发布，包含具体的版本号和性能细节。
AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →
<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://pub.towardsai.net/claude-opus-4-8-is-4x-more-honest-that-honesty-is-eating-your-context-window-7d2f78bc4a08?source=rss------claude-5"><img src="https://cdn-images-1.medium.com/max/1376/1*TosrbBHROro3UFkI2…
<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@elenagod243/claude-opus-4-8-is-here-what-making-honesty-a-feature-actually-means-7024dad7821f?source=rss------anthropic-5"><img src="https://cdn-images-1.medium.com/max/2600/1*YTHPesYd9uZSKFfg…
<div class="medium-feed-item"><p class="medium-feed-snippet">Anthropic shipped Claude Opus 4.8 on May 28, 2026. The benchmark numbers are impressive — SWE-bench Pro up to 69.2%, USAMO 2026 at 96.7%…</p><p class="medium-feed-link"><a href="https://medium.com/@kosukeo…
<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://ai.plainenglish.io/claude-opus-4-8-the-quiet-turn-toward-honesty-30bd2192c795?source=rss------claude-5"><img src="https://cdn-images-1.medium.com/max/2600/1*c2U58Vk6iGOQ87xO-kQTyQ.png" width="2816" /></a>…
<p><strong>TL;DR —</strong> On <strong>May 28, 2026</strong>, Anthropic released <strong>Claude Opus 4.8</strong> (<code>claude-opus-4-8</code>), the second Opus upgrade in under two months. Headline numbers: <strong>SWE-Bench Pro 64.3% → 69.2%</strong>, <strong>4× less likely to…