Perplexity's new contextual embedding model tops turbopuffer's context-bench with 50%+ recall gain
marktenenholtz · x · 2026-10-01
turbopuffer reports that Perplexity's new model tops context-bench, its internal contextual embedding benchmark, boosting document recall@10 by 50%+ over traditional SOTA embedding models. marktenenholtz amplified the news with a cheeky "Are you entertained?" turbopuffer says it maintains internal benchmarks to guide customers toward better search relevance.
More from Models
- Accusation: Google Deliberately Omitted Sol, Opus and Fable from Its Benchmark Table — zacharynado · 2026-10-01
- TheZvi Jokes About What Horrors Might Lurk Beneath Gemini 4 Argon — TheZvi · 2026-10-01
- DeepSWE v1.1: Gemini 4 Argon Edges Out Claude Opus 5.5 and GPT-6 Astra — rohanpaul_ai · 2026-10-01
- Researchers Congratulate Google as Argon Signals a Return to the Frontier — andrew_n_carr · 2026-10-01
- GDP.xlsx benchmark: 70 real spreadsheet tasks, best frontier agent scores only 38.3% — echen · 2026-10-01
- Gemini 4 Argon beats quantum computing baseline by 40% in minutes, says Google researcher — LinusEkenstam · 2026-10-01