Close Menu
  • AI
  • Content Creation
  • Tech
  • Robotics
AI-trends.todayAI-trends.today
  • AI
  • Content Creation
  • Tech
  • Robotics
Trending
  • Potential Cognition AI Spherical Would Elevate $2B-Plus at $75B Valuation – Unite.AI
  • OpenRouter Cuts Three Suppliers After AI Immediate Publicity Findings – Unite.AI
  • A Group of Researchers from Columbia and CertiK Introduce TensorLean for Formal Verification of GPU Kernels
  • Collectively AI Discloses Immediate Leak After Coding Agent Uncovered Credential – Unite.AI
  • Dragos Finds Immediate Leaks and Stolen Keys in AI Provide Chains
  • Lean and Mathlib Communities Certify Claude-Assisted Riemann Proof – Unite.AI
  • AWS Routes HyperPod Inference Throughout Areas From One Endpoint – Unite.AI
  • DailyAI Briefing: A Lean-checked Riemann proof, plus 4 extra
AI-trends.todayAI-trends.today
Home»Tech»Liquid AI Introduces LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M: Dense Bi-Encoder and Late-Interplay Fashions for Quick Multilingual Search Throughout 11 Languages

Liquid AI Introduces LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M: Dense Bi-Encoder and Late-Interplay Fashions for Quick Multilingual Search Throughout 11 Languages

Tech By Gavin Wallace19/06/20266 Mins Read
Facebook Twitter LinkedIn Email
A Coding Implementation to Build an Interactive Transcript and PDF
A Coding Implementation to Build an Interactive Transcript and PDF
Share
Facebook Twitter LinkedIn Email

This week, Liquid AI launched two new retrieval fashions. They’re LFM2.5-ColBERT-350M and LFM2.5-Embedding-350M. Each maintain 350M parameters. Each are the primary bidirectional members of the LFM household. They construct on LFM2.5-350M-Base, launched in March. The pair targets quick multilingual and cross-lingual search throughout 11 languages. Their footprint is sufficiently small to run virtually wherever. Each can be found now on Hugging Face beneath the LFM Open License v1.0.

LFM2.5 Retrievers

The 2 fashions share one spine however signify textual content in a different way. LFM2.5-Embedding-350M is a dense bi-encoder. It turns every doc right into a single vector. Decide it once you need the quickest search and the smallest, most cost-effective index.

LFM2.5-ColBERT-350M is a late-interaction mannequin. It converts every token right into a vector moderately than one vector per doc. This lets it match queries word-by-word for greater accuracy and higher generalization. The trade-off is a bigger index. Decide it when accuracy issues greater than storage. Its question size is capped at 32 tokens. It could possibly additionally rerank a first-stage retriever’s outcomes with out constructing an index.

Each goal short-context search. Good matches embody product catalogs, FAQ data bases, and help docs. Liquid AI positions each as a drop-in substitute for an current RAG pipeline.

The Structure Change: Causal to Bidirectional

Each fashions begin from LFM2.5-350M-Base, a mid-trained general-purpose checkpoint. Liquid AI applies a small set of bidirectional patches to the LFM2 structure. These adapt it from a causal decoder to a bidirectional encoder.

In a causal setup, every token makes use of solely itself and former tokens. That fits left-to-right era however is much less pure for retrieval. The staff replaces the causal consideration masks with a bidirectional one. Now each token can attend to each left and proper context. In addition they make the LFM2 brief convolutions non-causal. These combine native info symmetrically round every token, not solely from the previous.

This preserves the LFM2 spine’s effectivity whereas producing the full-context representations retrieval wants. Every mannequin has 17 layers: 10 convolution, 6 consideration, and 1 pooling or dense. Context size reaches 32,768 tokens, although paperwork are tuned to 512 tokens. From the shared encoder, the 2 fashions differ solely in output. Embedding makes use of CLS-style pooling for one 1024-dim vector. ColBERT retains 128-dim per-token embeddings for MaxSim late interplay.

Coaching and Information

Each fashions comply with the identical three-stage recipe:

  • Stage one is large-scale contrastive pretraining in English.
  • Stage two is multilingual and cross-lingual distillation from a powerful instructor throughout all 11 languages.
  • Stage three is closing fine-tuning on hard-mined negatives.

The Embedding mannequin receives barely extra cross-lingual information than ColBERT. Cross-lingual retrieval emerges extra naturally within the late-interaction setup. Coaching information combines curated inner information with open-source English retrieval datasets. LLM-based translation expands the multilingual and cross-lingual pairs.

Benchmark

Liquid AI evaluated two capabilities. The primary is multilingual retrieval with NanoBEIR. The second is cross-lingual open-domain QA with MKQA-11. Each report outcomes throughout all 11 languages: Arabic, German, English, Spanish, French, Italian, Japanese, Korean, Norwegian, Portuguese, and Swedish.

On common, each fashions lead their class. Listed below are the comparability particulars:

Mannequin Kind NanoBEIR ML (NDCG@10) MKQA-11 (Recall@20)
LFM2.5-ColBERT-350M late interplay 0.605 0.694
LFM2.5-Embedding-350M dense 0.577 0.691
Qwen/Qwen3-Embedding-0.6B dense 0.556 0.638
LFM2-ColBERT-350M late interplay 0.540 0.646
Alibaba-NLP/gte-multilingual-base dense 0.528 0.675
lightonai/GTE-ModernColBERT-v1 late interplay 0.489 0.459
BAAI/bge-large-en-v1.5 dense 0.359 0.413

ColBERT leads on each averages. Embedding is shut behind on MKQA-11 at 0.691. Each beat Qwen3-Embedding-0.6B, a bigger mannequin. The brand new ColBERT additionally improves on the sooner LFM2-ColBERT-350M, from 0.540 to 0.605 on NanoBEIR. Liquid AI additionally notes that NanoBEIR English tracks the costlier full BEIR. The 2 keep extremely correlated, with NanoBEIR scoring a near-constant ~15% greater. The analysis staff subsequently makes use of NanoBEIR as a sensible proxy throughout coaching runs.

Latency and Edge Deployment

Liquid AI launched GGUF variants for llama.cpp. These let each fashions run on CPUs, laptops, and edge gadgets. The figures under use a MacBook Professional M4 Max at FP16. Queries are 32 tokens; paperwork are 256 tokens.

Mannequin Stage Docs cached p50
LFM2.5-Embedding-350M Question embedding sure 7.3 ms
LFM2.5-ColBERT-350M Question embedding + MaxSim sure 8.2 ms
LFM2.5-ColBERT-350M Question + Doc embedding + MaxSim no 34.3 ms

When doc embeddings are pre-computed, median (p50) question latency stays beneath 10 ms. Encoding paperwork at question time pushes ColBERT to 34.3 ms. For enterprise scale, Liquid AI additionally constructed an inner GPU stack. On an H100 at FP16, it observes latencies as little as 1 ms. Embedding question latency there’s 1.5 ms p50.

Use Circumstances With Examples

  • E-commerce: Search a product catalog throughout many languages with one index. A client varieties a Korean question and the system surfaces an English product itemizing. Cross-lingual retrieval makes this work with out per-language indexes.
  • FAQ and help data bases: Retrieve the correct reply reliably throughout customer-facing surfaces. A French help query maps to an English assist article.
  • On-device semantic search: Search recordsdata, emails, and notes regionally on client {hardware}. The GGUF construct retains information on the gadget at near-zero price.
  • Enterprise data assistants: Retrieve inner authorized, monetary, and technical paperwork throughout languages. ColBERT fits this when reply accuracy outranks index dimension.

Code: Getting Began

The Embedding mannequin runs by sentence-transformers. All the time go the uneven prompts, question: and doc:. Omitting them silently degrades retrieval high quality.

from sentence_transformers import SentenceTransformer

mannequin = SentenceTransformer(
    "LiquidAI/LFM2.5-Embedding-350M",
    trust_remote_code=True,
)

queries = ["What is the capital of France?"]
paperwork = ["Paris is the capital and largest city of France."]

q_emb = mannequin.encode(queries,   prompt_name="query",    normalize_embeddings=True)
d_emb = mannequin.encode(paperwork, prompt_name="document", normalize_embeddings=True)

scores = q_emb @ d_emb.T  # form: (n_queries, n_documents)

The ColBERT mannequin runs by PyLate. Its PLAID index makes use of FastPLAID for environment friendly similarity search.

from pylate import indexes, fashions, retrieve

mannequin = fashions.ColBERT(
    model_name_or_path="LiquidAI/LFM2.5-ColBERT-350M",
    trust_remote_code=True,
)
mannequin.tokenizer.pad_token = mannequin.tokenizer.eos_token

index = indexes.PLAID(index_folder="pylate-index", index_name="index", override=True)

docs_emb = mannequin.encode(["document 1 text", "document 2 text"], is_query=False)
index.add_documents(documents_ids=["1", "2"], documents_embeddings=docs_emb)

retriever = retrieve.ColBERT(index=index)
q_emb = mannequin.encode(["a search query"], is_query=True)
scores = retriever.retrieve(queries_embeddings=q_emb, ok=10)

To rerank an current first-stage pipeline as an alternative, skip the index and use rank.rerank.

from pylate import fashions, rank

mannequin = fashions.ColBERT(model_name_or_path="LiquidAI/LFM2.5-ColBERT-350M", trust_remote_code=True)

queries = ["query A"]
paperwork = [["candidate doc 1", "candidate doc 2"]]
documents_ids = [[1, 2]]

q_emb = mannequin.encode(queries, is_query=True)
d_emb = mannequin.encode(paperwork, is_query=False)

reranked = rank.rerank(
    documents_ids=documents_ids,
    queries_embeddings=q_emb,
    documents_embeddings=d_emb,
)

You can even fine-tune both mannequin by yourself information. The Embedding card offers snippets utilizing sentence-transformers and MultipleNegativesRankingLoss.

Key Takeaways

  • Liquid AI’s LFM2.5-ColBERT-350M and LFM2.5-Embedding-350M are the primary bidirectional LFMs, constructed for multilingual search throughout 11 languages.
  • Each 350M fashions lead their class on NanoBEIR and MKQA-11, beating the bigger Qwen3-Embedding-0.6B.
  • Embedding provides the smallest, most cost-effective index; ColBERT trades a bigger index for greater per-token accuracy.
  • GGUF builds run on CPUs, laptops, and edge by way of llama.cpp, with cached p50 question latency beneath 10 ms.
  • They drop into current RAG pipelines by sentence-transformers and PyLate, beneath the LFM Open License v1.0.

Interactive Explainer


AI ar ces code fashion languages search
Share. Facebook Twitter LinkedIn Email
Avatar
Gavin Wallace

Related Posts

A Group of Researchers from Columbia and CertiK Introduce TensorLean for Formal Verification of GPU Kernels

22/11/2026

OpenAI Ships GPT-Stay-1.1 for Voice Brokers, Cuts Idle Value to $0.02/Min

21/11/2026

Meet Pondering Machines Inkling 2 Mini: 26B Looped MoE Runs on a 32GB Mac

21/11/2026

Baseten Releases Bison-1 and Groundhog-1, Topping Choice Mannequin Index

21/11/2026
Top News

Hackers hijacked Google Gemini AI with a poisoned calendar invite to take over a smart home

Ransomware based on AI is now a reality

WIRED| WIRED

AI isn’t coming for Hollywood. It has already arrived

You can jailbreak some Frontier AI models with frightening ease

Load More
AI-Trends.Today

Your daily source of AI news and trends. Stay up to date with everything AI and automation!

X (Twitter) Instagram
Top Insights

Context Graphs – What Are They? MarkTechPost

21/01/2026

This article provides a comprehensive review, comparative insights and use cases for the best Chinese open agentic/reasoning models (2025).

11/08/2025
Latest News

Potential Cognition AI Spherical Would Elevate $2B-Plus at $75B Valuation – Unite.AI

22/11/2026

OpenRouter Cuts Three Suppliers After AI Immediate Publicity Findings – Unite.AI

22/11/2026
X (Twitter) Instagram
  • Privacy Policy
  • Contact Us
  • Terms and Conditions
© 2026 AI-Trends.Today

Type above and press Enter to search. Press Esc to cancel.