Anthropic发布Haiku 5.5:成本降90%
Anthropic dropped Haiku 5.5
Haiku系列迎来重大迭代,成本腰斩且智能体能力显著增强,对Agent开发者的性价比选型有直接影响,值得关注。
Anthropic dropped Haiku 5.5
Anthropic 发布了 Haiku 5.5
> costs 90% less than Haiku 4.5 on prompts up to 100K tokens.
> 在处理多达 100K token 的提示词时,其成本比 Haiku 4.5 低 90%。
> Haiku 5.5 adds the first Haiku effort setting, trading cost for accuracy, but Anthropic still recommends Sonnet 5.5 and Opus 5.5 for complex agentic coding such as Terminal-Bench 4.0 tasks.
> Haiku 5.5 新增了首个 Haiku 努力程度设置,以成本换取准确性,但 Anthropic 仍建议在终端基准测试 4.0 等复杂的代理式编码任务中使用 Sonnet 5.5 和 Opus 5.5。
> Asana reported over 30% lower task-completion latency and up to 2.5x faster inference per agent turn than its current model.
> Asana 报告称,与当前模型相比,其任务完成延迟降低了 30% 以上,且每个代理回合的推理速度提高了 2.5 倍。
> On OSWorld 2.1, which tests whether an AI agent can operate a real computer to finish long multi-step tasks, Haiku 5.5 jumped from Haiku 4.5's 15.7% to 72.4%.
> 在 OSWorld 2.1(用于测试 AI 代理能否操作真实计算机以完成长多步任务的基准)上,Haiku 5.5 的表现从 Haiku 4.5 的 15.7% 跃升至 72.4%。
On Chartography, a visual reasoning test of reading and interpreting charts without tools, it rose from 6.4% to 46.4%.
在 Chartography(一项无需工具即可阅读和解读图表的视觉推理测试)中,其得分从 6.4% 上升至 46.4%。
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力