Topic
Techniques for controlling what an LLM keeps in its context window, such as compaction, summarization and elision, which shape token cost and usage limits.
Anthropic switched on idle compaction in Claude Code around v2.1.286, replacing long idle conversations with a summary, a dev.to analysis reports. Version 2.1.290 adds idleCompaction: false, a choice that falls on anyone keeping large sessions open past an hour.
Reality
Researchers held a coding harness's execution loop fixed and varied planning, tools and context management across four models, and found that what each component is worth depends on how strong the model already is.
Reality
The 35-billion-parameter Iris-mini and the 397-billion-parameter Iris-pro build on Qwen models and run at 256,000 tokens of context. AllSpark also published the training recipe and scored every benchmark twice, once with context management switched off.
Reality
Token use has outgrown the accounting around it, and the firms reporting returns are the ones putting the price of a prompt in front of the person about to run it, rather than in a dashboard read a quarter later.
Reality
Anthropic meters Pro and Max on a rolling 5-hour window and a weekly cap, both counted in tokens. What drains them is re-read context, not how many questions you ask.
Reality