GGUF format of Alibaba-NLP/gte-base-en-v1.5 for use with CrispEmbed.
GTE Base EN v1.5. Post-LN BERT with NTK-scaled RoPE and GeGLU. 768-dimensional output, CLS pooling. 8192-token context.
| File | Quantization | Size |
|---|---|---|
| gte-base-en-v1.5.gguf | F32 | 522 MB |
| gte-base-en-v1.5-q8_0.gguf | Q8_0 | 139 MB |
| gte-base-en-v1.5-q4_k.gguf | Q4_K | 85 MB |
Cosine similarity vs the upstream sentence-transformers reference on a fixed test set (text):
| Quant | Text |
|---|---|
| q8_0 | 0.9836 |
| q4_k | 0.9178 |
Note: below the 0.99 retrieval-quality bar — text: f32 (0.984), q8_0 (0.984), q4_k (0.918). Embeddings are still functionally usable (>0.9 = directionally correct for similarity ranking) but expect small differences in nearest-neighbor results vs the upstream f32 reference.
# Download
huggingface-cli download cstr/gte-base-en-v1.5-GGUF gte-base-en-v1.5.gguf --local-dir .
# Run with CrispEmbed
./crispembed -m gte-base-en-v1.5.gguf "Hello world"
# Or with auto-download
./crispembed -m gte-base-en-v1.5 "Hello world"
| Property | Value |
|---|---|
| Architecture | GTE v1.5 (New BERT) |
| Parameters | 137M |
| Embedding Dimension | 768 |
| Layers | 12 |
| Pooling | CLS |
| Tokenizer | WordPiece |
| Base Model | Alibaba-NLP/gte-base-en-v1.5 |
Verified bit-identical to HuggingFace sentence-transformers (cosine similarity >= 0.999 on test texts).
CrispEmbed is a lightweight C/C++ text embedding inference engine using ggml. No Python runtime, no ONNX. Supports BERT, XLM-R, Qwen3, and Gemma3 architectures.
# Build CrispEmbed
git clone https://github.com/CrispStrobe/CrispEmbed
cd CrispEmbed
cmake -S . -B build && cmake --build build -j
# Encode
./build/crispembed -m gte-base-en-v1.5.gguf "query text"
# Server mode
./build/crispembed-server -m gte-base-en-v1.5.gguf --port 8080
curl -X POST http://localhost:8080/v1/embeddings \
-d '{"input": ["Hello world"], "model": "gte-base-en-v1.5"}'
convert-bert-embed-to-gguf.pyAlibaba-NLP.apache-2.0. This repository redistributes under the same terms; it grants no rights the upstream licence does not.Alibaba-NLP/gte-base-en-v1.5