3 ms·

DeepSeek’s docs say V4 has a 1M context length. Is that actually usable in practice, or just the model/API limit? Codex shows ~258k for me and Claude Code ofte

by Grp1 5mo ago

DeepSeek’s docs say V4 has a 1M context length. Is that actually usable in practice, or just the model/API limit? Codex shows ~258k for me and Claude Code often shows ~200k, so I’m curious how DeepSeek is exposing such a large window.

lucrbvi 5mo ago
They have added a lot of optimization focussing on the KV-cache, so they can have a much larger window without eating all the VRAM. The 1M window might be usable, but it will probably underperform against a smaller window of course.
- Grp1 5mo ago
  [dead]