Intelligence cost-effectiveness
Original title:Intelligence Per Dollar
AI summary
Microsoft added an average token usage metric to its model release card for the first time. Its model scores 71.6 on SWE-Bench Verified while consuming only about one-third the tokens of Claude Haiku 4.5. Artificial Analysis's Intelligence Index shows GPT 5.5 and Claude Opus 4.8 scoring similarly (about 60 points), but Opus 4.8 costs 40% more to run ($4,685 vs $3,357). Uber limited employee usage after its AI budget overran within four months; Salesforce spent $300 million on Anthropic tokens and froze engineering hiring. Model companies now must compete on both performance and cost.
Key Takeaways
- Microsoft added an average token usage metric to its model release card for the first time.
- Its model scores 71.
- 6 on SWE-Bench Verified while consuming only about one-third the tokens of Claude Haiku 4.
- Artificial Analysis's Intelligence Index shows GPT 5.
- 5 and Claude Opus 4.
Quick Facts
- Published
- Updated
- Source
- Tomasz Tunguz 博客(VC 分析)
- Category
- Opinion
- AI score
- 66
微软在模型发布卡上悄悄加了“平均token消耗”这个指标,这不是小改动,而是宣告AI从堆算力转向算账时代。Uber和Salesforce的预算教训已经很清楚了。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Intelligence cost-effectiveness. June 3, 2026. https://new.nat6.net/en/news/cmpy42n0t01jnslaxhzgxjmzh/