A display-only multilingual QA retrieval benchmark reported by Liquid AI for LFM2.5 retriever models, using Recall@20 across 11 languages.
BenchLM mirrors the published score view for MKQA-11. LFM2.5-ColBERT-350M leads the public snapshot at 69.4% , followed by LFM2.5-Embedding-350M (69.1%). BenchLM does not use these results to rank models overall.
LFM2.5-ColBERT-350M
LiquidAI
LFM2.5-Embedding-350M
LiquidAI
Year
2026
Tasks
Cross-lingual open-domain QA retrieval
Format
Recall@20 average
Difficulty
Multilingual retrieval
Liquid reports MKQA-11 average Recall@20 across Arabic, German, English, Spanish, French, Italian, Japanese, Korean, Norwegian, Portuguese, and Swedish. BenchLM stores the average as a display-only retrieval signal.
Version
MKQA-11 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
A display-only multilingual QA retrieval benchmark reported by Liquid AI for LFM2.5 retriever models, using Recall@20 across 11 languages.
LFM2.5-ColBERT-350M by LiquidAI currently leads with a score of 69.4% on MKQA-11.
2 AI models have been evaluated on MKQA-11 on BenchLM.
For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.