gte-modernbert-base-mlx-8bit
8-bit integer affine quantization of the embedder's linear weights (group size 64), produced with MLX. CLS pooling as the source's 1_Pooling/config.json. Measured against the float16 source on the Premisa retrieval study's chunk set (180 questions, 50-candidate pools): 98 percent top-8 neighbour overlap; the full study arm at 2000-character chunks and pool 50 matched the source's delivered-hit figures. English-only, as the source.
Source: Alibaba-NLP/gte-modernbert-base at revision e7f32e3c00f91d699e8c43b53106206bcc72bb22 (source weights SHA-256 3e85899d5728cb7de79781c0c3acfb91ccef9f875f1f7e0b3c9f3dd4b6a724ba), quantized on 2026-10-21 with Premisa's model_quant_spike.py quantize --bits 8 (mode affine, group size 64). Serving: the mlx-vlm fork's task server (python -m mlx_vlm.server --embedding-model or --reranker-model with this directory); the pooling configuration is read from 1_Pooling/config.json where present.
License: the source model's license applies to these derived weights (apache-2.0); see the source card for its terms. Quantization and measurements by Premisa (ai.premisa.desktop); figures are pre-adjudication study measurements, not quality claims.