--- base_model: microsoft/harrier-oss-v1-0.6b library_name: transformers.js license: mit language: [multilingual] tags: [embedding, onnx, int8, transformers.js, harrier, multilingual, qwen3] pipeline_tag: feature-extraction --- # Harrier-OSS-v1-0.6B — ONNX INT8 (single-file) Re-export of [microsoft/harrier-oss-v1-0.6b](https://huggingface.co/microsoft/harrier-oss-v1-0.6b) as single-file ONNX INT8 for Transformers.js v4 and ONNX Runtime. | Property | Value | |----------|-------| | Dimensions | **1024** (same as pplx-embed-v1) | | MTEB v2 | **69.0** | | Languages | 94 | | Max tokens | 32,768 | | Pooling | Last-token + L2 norm | | Architecture | Qwen3 (vocab 151,936) | | Size | 600 MB (INT8) | | License | MIT | **Important**: Queries must be prefixed with an instruction. See `search_instructions.json`. Potential **drop-in replacement** for pplx-embed-v1 (same 1024D, same vocab family). Export by [Deposium](https://deposium.ai).