#model evaluation

model evaluation

2 篇论文

A librarian and two adult students sort resource cards and check missing catalogue fields while recording metadata notes.

政策 / 伦理2026年10月9日

Product news: Claude Haiku 5.5 expands fast, small-model options for narrowly defined educational tasks

Anthropic

AI Product and Learning Report

Anthropic released Claude Haiku 5.5 on October 7 for fast, high-volume work and added an adjustable effort setting. This October 9 report examines bounded educational uses such as organising public resource metadata. The same announcement lowers Sonnet 5.5 cache-read pricing. OpenAI's model guide and Gemini's learning context provide comparisons. Lower cost can broaden experimentation, but each task still needs checks for omissions, incorrect categories, and cases requiring a person.

product newsClaude Haiku 5.5small models

阅读 500 字摘要 →

A lecturer and postgraduate compare two anonymous response sheets with a source table while another student records review effort.

政策 / 伦理2026年9月23日

政策 / 伦理 173

Product news: GPT-6 Sol and Luna broaden lower-cost model choices for educational and research workflows

OpenAI

AI Product and Learning Report

OpenAI released GPT-6 Sol and Luna on September 22, extending its model family with lower-priced options for coding, professional work, and computer use. This September 23 report focuses on choosing a model for a specific educational task. Anthropic's same-day Opus 5.5 release and Gemini's inquiry tools provide context. Token prices matter, but a useful comparison also includes errors, retries, review effort, and what learners are expected to do themselves.

product newsGPT-6 SolGPT-6 Luna

阅读 500 字摘要 →