Opus Dominates Long-Context Retrieval
Long-context retrieval MRCR v2 (8-needle) MEAN MATCH RATIO (%) 93.0 76.0 10.8 18.5 Opus 4.6 256k Opus 4.6 1M Sonnet 4.5 256k Sonnet 4.5 1M
A bar chart compares long-context retrieval performance for Opus 4.6 and Sonnet 4.5 at 256k and 1M context lengths. The joke is that Opus dramatically outperforms Sonnet, making the prospect of a larger context window and improved retrieval especially appealing to AI enthusiasts.
- @dev_meme / 7676 2026-02-05
- Type
- image
- Size
- 800×450 · 26.2 KiB
- Mood
- sarcastic
- Language
- en
- Added
