| 比 Opus 4.6 更高分辨率的图像输入 |
来自 Anthropic 的发布帖子。它的基准比较是图像,因此没有保存基准图表(差距);帖子文本中未说明启动上下文窗口图。 · 来源
Estimate from the version's published specs, benchmarks and reviews, 0-5; scores are only comparable within one criteria year. Scored as launched (Claude Opus 4.7, 16 Apr 2026): Artificial Analysis Intelligence Index 57 (57.3 exact; max effort), tied with Gemini 3.1 Pro and GPT-5.4, per its 17 Apr 2026 article (index version then current, rebased since). Coding null: no Terminal-Bench 4.0 result at launch (Anthropic cites Terminal-Bench 2.0). 1M-token context; $5/$25 per 1M (about $10 blended), though a new tokenizer maps the same text to roughly 1.0-1.35x more tokens. 54.5 output tokens/s on Artificial Analysis (live page, 24 Oct 2026). Text and image input (about 3.75 MP). 4 reviews, string-matched.[1]
Claude Opus 4.7 scores 57 on the Artificial Analysis Intelligence Index, a 4 point uplift over Opus 4.6 (Adaptive Reasoning, Max Effort, 53).
For factual hallucinations, Opus 4.7 performs better than or on par with Opus 4.6 across four benchmarks but falls short of Mythos Preview.
In return, users get slightly better instruction following: a test using the IFEval benchmark across 20 prompts shows Opus 4.7 sticking to strict instructions five percentage points more reliably than its predecessor.
Claude not only gave a better and more thorough response, but elevated the answer to a professional/research level.
读者上传的照片,与图钉带来源的媒体分开显示。你可以为吃到或看到的内容评分,在餐厅图钉上还能直接在照片上标记菜品。
第一项始终是图钉的来源。总体可信度是各份资料对上方所用开始和结束时间支持程度的加权平均;资料每比最新的一份旧 180 天,权重减半。
来自 Anthropic[3] 官方公告,日期即模型正式可用当天。
Anthropic's[1][3] Claude Opus 4.7 model page lists a 1M-token context window, 128K max output tokens, adaptive thinking, text and images to text, and a Jan 2026 reliable knowledge cutoff.
Anthropic's[1] Claude[2] Opus page lists this announcement and its date.
有遗漏或错误吗?用你自己的话说明:能佐证此图钉的链接、不同的开始或结束日期及理由,或缺失、有误的信息。AI 会对照此图钉的来源进行核实,搜索更好的来源,并添加任何支持你说法的页面。图钉自身的来源仍然最重要。AI 也会查看图片:显示的是别的东西或显示效果差的图片会被移到后面或替换。
原标题: Claude Opus 4.7 Released