In depth
The context window caps how much input plus output fits in one call. Claude Sonnet ships 200K, GPT-4o is 128K, Gemini Pro pushes to 2M. Long sessions overflow eventually, which is why compaction exists. Bigger windows are not always better: cost scales with input tokens, and models often degrade in quality at the far edge of their window.
Related concepts
TokensThe units LLMs process. Roughly four characters of English per token, billed per million.CompactionA runtime process that summarises old turns once the context window is nearly full, freeing space for new turns.Frontier modelThe highest-quality, most expensive tier from a provider. Claude Sonnet, GPT-4o, Gemini Pro.
Newsletter
Get the next post in your inbox.
Engineering notes from the Digitorn team. No marketing, no launch announcements, no "10 prompts that will change your life". Just the things we write that we'd want to read.