Anthropic engineers on model upgrades, caching and cost per task

Upgrading models is a cost decision as much as a quality one. This explains how caching and mid-conversation effort changes affect spend, and why to re-tune your harness and track cost per task.

• prompt caching — Reusing the model's processed form of a repeated prompt prefix so subsequent calls skip re-reading it, cutting cost and latency substantially.
• agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.

• postAnthropic gives all paid Claude users a one-time full limit reset
• articleAnthropic's Guide to Choosing a Claude Model and Effort Level
• postFable 5.2 Ships Faster and 20% Cheaper Than 5.1
• postClaude Code Cloud Sessions Now Auto-Configure and Save Repo Environments

• postWhy the cheaper model sometimes costs more per taskWRITER
• articleHarness Design Cut Agent Cost 33-61% Across Six ModelsWRITER
• videoWithout Caching, Your AI Agent Costs 10x More - George Cameron, Artificial AnalysisAI Engineer

Checking sign-in…

Loading comments…