Model reference · open weights
LFM2-ColBERT is an open-weight embedding model from LiquidAI, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.
About
src="https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/2b08LKpev0DNEk6DlnWkY.png" alt="Liquid AI" style="width: 100%; max-width: 100%; height: auto; display: inline-block; margin-bottom: 0.5em; margin-top: 0.5em;" / LFM2-ColBERT-350M LFM2-ColBERT-350M is a late interaction retriever with excellent multilingual performance. It allows you to store documents in one language (for example, a product description in English) and retrieve them in many languages with high accuracy. - LFM2-ColBERT-350M offers best-in-class accuracy across different languages. - Inference speed is on par with models 2.3 times smaller, thanks to the efficient LFM2 backbone. - You can use it as a drop-in replacement in your current RAG pipelines to improve performance. Find more information about LFM2-ColBERT-350M in our blog post. [!NOTE] 🚀 Try our demo: https://huggingface.co/spaces/LiquidAI/LFM2-ColBERT 📄 Model details Late interaction retrievers like LFM2-ColBERT-350M are particularly interesting because they preserve much of the expressivity of re-rankers while retaining the efficiency of bi-encoders. In practice, they're used to both retrieve documents at scale (like bi-encoders) and rank them at the same time (like rerankers). We recommend using this model for various RAG use cases, such as: - E-commerce: Find products across many languages with semantic search at scale. - On-device semantic search: Ask questions to your phone in natural language to retrieve files, emails, and notes. - Enterprise knowledge assistants: Retrieve internal legal, financial, and technical documents in different languages. Document length: 512 tokens Query length: 32 tokens Output dimensionality: 128 tokens Similarity function: MaxSim Supported languages: English, Arabic, Chinese, French, German, Japanese, Korean, and Spanish. 🏃 How to run Sentence Transformers This model can be used with Sentence Transformers as a multi-vector (ColBERT-style late interaction) retriever via the MultiVectorEncoder: PyLate First, install the PyLate and transformers library: Retrieval Use this model with PyLate to index and retrieve documents. The index uses FastPLAID for efficient similarity searc
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | LiquidAI |
|---|---|
| Type | Embedding models |
| Parameters (lead) | 353M |
| Context | 125k tokens |
| Variants | 1 |
| Runs with | PyLate |
| Released | 2025-10-28 |
| Popularity | 25k downloads / month |
| Likes | 152 |
| Licence | Commercial licence needed |
How it works
Variants
Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.
| Variant | Params | Precision | VRAM | Fits 16 GB | Weights |
|---|---|---|---|---|---|
| LFM2-ColBERT-350M | 353M | BF16 | ~0.8 GB | ✓ | Weights ↗ |
Benchmarks
As published on the model card — the maker's own numbers, not measured by AxForge.
| Task | Dataset | Metric | Score |
|---|---|---|---|
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Accuracy@1 | 0.4 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Accuracy@3 | 0.52 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Accuracy@5 | 0.64 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Accuracy@10 | 0.8 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Precision@1 | 0.4 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Precision@3 | 0.207 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Precision@5 | 0.156 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Precision@10 | 0.118 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Recall@1 | 0.195 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Recall@3 | 0.273 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Recall@5 | 0.316 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Recall@10 | 0.459 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Ndcg@10 | 0.387 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Mrr@10 | 0.506 |
| Py Late Information Retrieval | NanoClimateFEVER | Maxsim Map@100 | 0.313 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Accuracy@1 | 0.86 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Accuracy@3 | 0.92 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Accuracy@5 | 0.94 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Accuracy@10 | 0.98 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Precision@1 | 0.86 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Precision@3 | 0.7 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Precision@5 | 0.64 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Precision@10 | 0.566 |
| Py Late Information Retrieval | NanoDBPedia | Maxsim Recall@1 | 0.121 |
Using it via the API
Once AxForge deploys lfm2-colbert for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (lfm2-colbert below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/embeddings \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"lfm2-colbert","input":"text to embed"}'
Licence
The weights are open but its licence needs a commercial agreement for business use. AxForge can arrange that licence and host the model for you — you pay AxForge, we settle with the model’s maker. Ask us for a quote. Read the licence ↗
Explore