LFM2.5 Encoder 230M Code MXFP4-GPTQ
This is a modified RESMP.DEV research release derived from
LiquidAI/LFM2.5-Encoder-230M at revision 0b649ad0c684378b03d4d8304f7577a662ab89bc. It is
not an official Liquid AI release. We removed the masked-language-model head and
contrastively fine-tuned the full bidirectional encoder for multilingual code retrieval.
Quantization finding
This is a research artifact, not an automatic recommendation to replace the BF16 model. Activation calibration is compared with matched native round-to-nearest quantization and the complete machine-readable receipts are included so mobile and Apple-Silicon users can evaluate the size, latency, memory, and quality tradeoff themselves.
Held-out retrieval results
All rows use the same untouched 6,995-pair multilingual test set, 1,200-character query and 4,000-character passage caps, query token cap 512, and passage token cap 2,048. Higher is better. RTN is a matched quantization control; Nomic and Jina are external service baselines, not architecture-matched controls.
| Model | MRR | R@1 | R@5 | R@10 | NDCG@10 | Python MRR | TypeScript MRR | Artifact |
|---|---|---|---|---|---|---|---|---|
| LFM2.5 230M BF16 | 0.3960 | 0.3212 | 0.4756 | 0.5375 | 0.4235 | 0.8360 | 0.2527 | 464.2 MB |
| LFM2.5 230M calibrated MXFP4 | 0.2110 | 0.1620 | 0.2572 | 0.3029 | 0.2262 | 0.7113 | 0.0693 | 225.5 MB |
| LFM2.5 230M RTN MXFP4 | 0.0806 | 0.0609 | 0.0931 | 0.1139 | 0.0844 | 0.4435 | 0.0191 | 225.4 MB |
| LFM2.5 230M calibrated MXFP8 | 0.3934 | 0.3197 | 0.4715 | 0.5342 | 0.4207 | 0.8394 | 0.2447 | 306.8 MB |
| LFM2.5 230M RTN MXFP8 | 0.3878 | 0.3162 | 0.4622 | 0.5245 | 0.4139 | 0.8322 | 0.2459 | 306.7 MB |
| Nomic v1.5 service | 0.5439 | 0.4968 | 0.5954 | 0.6236 | 0.5595 | 0.9289 | 0.3617 | service |
| Jina calibrated MXFP4 | 0.6645 | 0.6133 | 0.7221 | 0.7571 | 0.6832 | 0.9462 | 0.5057 | 1167.7 MB |
A separate BF16 cross-runtime run on NVIDIA GeForce RTX 3090 Ti with PyTorch 2.13.0+cu130 produced MRR 0.3963, 755.7 queries/s, 190.8 passages/s, and 736.2 MB peak CUDA allocation. CUDA throughput is reported separately and is not compared directly with Metal.
A paired 10,000-sample bootstrap estimates calibrated MXFP8 minus BF16 MRR at -0.0026, with a 95% interval of [-0.0042, -0.0010]. A point estimate whose interval crosses zero is not presented as a quality win.
Usage
git clone https://github.com/RESMP-DEV/calibrated-code-embeddings
cd calibrated-code-embeddings
uv sync --extra mlx
CODE_EMBEDDING_MODEL_PATH=/path/to/this-model code-embedding-serve --port 1235
The service exposes POST /v1/embeddings. It runs the bidirectional LFM2.5 body
directly with MLX; LM Studio is not required. Prefix retrieval queries with query:
and candidate code with passage: when calling the model directly.
Training and data receipts
Full-backbone symmetric in-batch InfoNCE training used 24,626 language-balanced pairs selected from the 42,626-row source training split, two epochs, batch size 32, learning rate 2e-5, temperature 0.05, and seed 17. The training report records the NVIDIA RTX A6000 runtime and validation history.
train: 42,626 rows, SHA-256426ebfaad34b14d7627ba6e668ae36e08e548c9d057b0edc208bcfa6fe527629validation: 5,319 rows, SHA-2569ac88b3138de4ca94c2ef3a87ccf19381fc76265c2bc9d65b4791983d0315096test: 6,995 rows, SHA-2569ed10842a12132b6bfb5421df1e2f88dbcfbf6f6e960f36b22eb9ea6e3c72315calibration: 4,096 rows, SHA-256ee9edaf80a6854c18053b96521090a51bdb76642abeb98618d7aed36e70b6de9
The corpus combines pinned CodeSearchNet data with pinned permissively licensed code
repositories. Exact and token 8-gram near-duplicates were removed with test-before-
validation-before-train precedence. See corpus_receipt.json, source_receipt.json,
training_report.json, quantization_report.json when present, benchmarks/, and
artifact_manifest.json for machine-readable evidence.
License and attribution
The weights retain the LFM Open License v1.0 in LICENSE, including its attribution and
commercial-use conditions. MODIFICATIONS.md identifies RESMP.DEV's changes. The
training and quantization workbench
is separately MIT licensed.
Citation
@article{liquidAI2026Encoders,
author = {Liquid AI},
title = {LFM2.5-Encoders: Fast at Long Context, Even on CPU},
journal = {Liquid AI Blog},
year = {2026},
note = {www.liquid.ai/blog/lfm2-5-encoders},
}