isitsota / categories / Embeddings
Embeddings / retrieval · Oct 13, 2026
Who is SOTA in Embeddings?
Microsoft Harrier Oss V1 27B & Tencent Kalm Embedding Gemma3 12B 2511 & Qwen Qwen3 Embedding 8B
Retrieval and embedding quality on the official MTEB multilingual leaderboard. English v2 is secondary and drops out if it saturates. As of Oct 13, 2026, Microsoft Harrier Oss V1 27B & Tencent Kalm Embedding Gemma3 12B 2511 & Qwen Qwen3 Embedding 8B share Embeddings. Microsoft Harrier…
Standings · overall
The board
| # | Model | State | Confidence |
|---|---|---|---|
| 1 | MSMicrosoft Harrier Oss V1 27B Microsoft | Co-SOTA | medium |
| 2 | TETencent Kalm Embedding Gemma3 12B 2511 tencent | Co-SOTA | medium |
| 3 | QQwen Qwen3 Embedding 8B Alibaba / Qwen | Co-SOTA | medium |
| 4 | BBytedance Seed1 6 Embedding 1215 ByteDance | Near-SOTA | medium |
| 5 | NNvidia Llama Embed Nemotron 8B NVIDIA | Near-SOTA | medium |
| 6 | QQwen Qwen3 Embedding 4B Alibaba / Qwen | Near-SOTA | medium |
| 7 | MSMicrosoft Harrier Oss V1.0 6B Microsoft | Near-SOTA | medium |
| 8 | COCodefuse AI F2llm V2 14B codefuse-ai | Near-SOTA | medium |
| 9 | GGoogle Gemini Embedding 001 Google DeepMind | Near-SOTA | medium |
| 10 | COCodefuse AI F2llm V2 8B codefuse-ai | Near-SOTA | medium |
| 11 | OCOcten Octen Embedding 8B Octen | Near-SOTA | medium |
| 12 | COCodefuse AI F2llm V2 4B codefuse-ai | Near-SOTA | medium |
| 13 | JIJinaai Jina Embeddings V5 Omni Small jinaai | Near-SOTA | medium |
| 14 | MSMicrosoft Harrier Oss V1 270m Microsoft | Not SOTA · #14 | medium |
| 15 | AIAI Sage Giga Embeddings Instruct 10B A1 8B 0826 ai-sage | Not SOTA · #15 | medium |
Evidence recipe
What can move this crown
- MTEB Multilingual v2 (MMTEB) — primary, weight 1
- MTEB English v2 — secondary, weight 0.45 (display only, cannot crown)