Total Cost of Agentic Tasks is Rising
ArtificialAnlys · x · 2026-07-14
At the AI Engineer World’s Fair, Artificial Anlys shared a key insight: 固定水平的“智能”正在变便宜，但完成复杂智能体任务的总成本反而在上升.
This is due to factors beyond just token price:
- 多轮交互: Agentic tasks often require multiple turns, and previous outputs feed back into the context, causing costs to accumulate.
- token 效率差异: Different models consume varying amounts of tokens for the same task; even with a lower unit price, the overall cost might be higher.
- prompt caching 命中率: Long tasks repeatedly pass the same instructions, research materials, and tool outputs, making cache hits increasingly impactful on the total price.
To profile such workloads, they introduced the AA-Briefcase benchmark to measure model performance in 长周期知识工作. Tasks span multi-week projects and thousands of input files, breaking down the costs of tool calls, turns, and tokens.
Related event: True AI Cost Depends on Task Completion Rates(2 posts)→
More from coding & agent
- AI agents are starting to strain code hosting platforms — craigsdennis · 2026-07-21
- Async OPD distillation doubles throughput while matching synchronous math accuracy — _lewtun · 2026-07-21
- Omnigent 0.6.0 adds Claude Code imports, Slack approvals and desktop apps — matei_zaharia · 2026-07-21
- Google appears to have quietly shipped Gemini 3.6 Flash, with lower pricing and better agentic scores — xiaohu · 2026-07-21
- Open-source CLI audits AI tools, MCP configs, and agent skills on local machines — Initial-Copy332 · 2026-07-21
- Coding agents feel less stressful when the 5-hour limits are temporarily removed — iamrobotbear · 2026-07-21