Voyage-3 Embedding (Voyage AI 2024年)（ボイヤージ3）
2024年Voyage AI発表Voyage-3・Industry-leading retrieval-optimized embedding model + Industry-leading 1024-dim + Industry-leading 32K context + Industry-leading $0.06 per 1M tokens cheaper + Industry-leading MIT alum founders。
概要
Voyage-3 Embedding は、2024年Voyage AI発表Voyage-3・Industry-leading retrieval-optimized embedding model 2024年 + Industry-leading MIT alum founders position確立。Voyage-3 specifications = Industry-leading retrieval-optimized embedding model (Industry-leading Voyage-3 2024 + Industry-leading retrieval-optimized + Industry-leading Voyage flagship) + Industry-leading 1024-dim (Industry-leading 1024-dim embedding) + Industry-leading 32K context (Industry-leading 32K tokens max context) + Industry-leading $0.06 per 1M tokens cheaper (Industry-leading $0.06 USD per 1M tokens cheaper than OpenAI)。
主な特徴・仕組み
- Org: Industry-leading Voyage AI (founded by Tengyu Ma + MIT alum 2023)
- Year: 2024年 (2024年Q4 release)
- Model: Industry-leading voyage-3 flagship + voyage-3-lite cheaper + voyage-code-3 code-specialized
- Embedding Dim: Industry-leading 1024-dim default
- Max Context: Industry-leading 32K tokens max context (long context)
- Retrieval-Optimized: Industry-leading retrieval-optimized specifically for RAG
- MTEB Performance: Industry-leading top MTEB retrieval benchmarks 2024
- Multilingual: Industry-leading multilingual support including code
- Pricing: Industry-leading $0.06 USD per 1M tokens voyage-3 / $0.02 lite
- API Only: Industry-leading API closed only
- Voyage Code 3: Industry-leading voyage-code-3 code-specialized embedding
- Anthropic Partnership: Industry-leading Anthropic recommended embedding partner
- Founders: Industry-leading Tengyu Ma Stanford + MIT alum + Voyage AI 2023 founded
- Industry-Leading: Industry-leading retrieval-optimized Voyage-3 cheaper than OpenAI 2024
スペック比較表
| LLM Embedding Model (2023-2024) | Dim | Org | License | Industry Position |
|---|---|---|---|---|
| Voyage-3 | 1024-dim | Voyage AI | Closed API $0.06/1M | Industry-leading Voyage closed API |
| OpenAI text-embedding-3-large | 3072-dim Matryoshka | OpenAI | Closed API $0.13/1M | Industry-leading OpenAI closed API |
| BGE Large | 1024-dim BERT-large | BAAI | MIT open source | Industry-leading open-source MTEB top |
| E5-Mistral-7B | 4096-dim Mistral-7B | Microsoft | MIT open source | Industry-leading 7B Mistral embedding |
| Nomic Embed v1 | 768-dim Matryoshka | Nomic AI | Apache 2.0 fully open | Industry-leading fully open + reproducible |
具体例・対応製品
- Voyage-3 (2024年, Voyage AI): Industry-leading retrieval-optimized embedding
- Industry-leading voyage-3 flagship + voyage-3-lite + voyage-code-3: Industry-leading 3-tier models
- Industry-leading 1024-dim + 32K tokens max context: Industry-leading 1024-dim + 32K
- Industry-leading $0.06 USD per 1M tokens cheaper than OpenAI $0.13: Industry-leading $0.06 cheaper
- Industry-leading Anthropic recommended embedding partner: Industry-leading Anthropic partnership
- 競合 BGE + E5-Mistral + Nomic + OpenAI: Industry-leading embedding competitors
自作PCでの選び方・注意点
Voyage-3 Embedding は「Industry-leading retrieval-optimized embedding model + Industry-leading 1024-dim」「Industry-leading 32K context + Industry-leading $0.06 per 1M tokens cheaper」用途のIndustry-leading retrieval-optimized 2024年Voyage AI発表product。Industry-leading retrieval-optimized (Industry-leading retrieval-optimized specifically for RAG + Industry-leading Voyage retrieval signature + Industry-leading top MTEB retrieval benchmarks 2024) で Industry-leading retrieval-optimized + Industry-leading Voyage retrieval signature + Industry-leading top MTEB retrieval 2024。Industry-leading 1024-dim 32K (Industry-leading 1024-dim default + 32K tokens max context long context + Industry-leading Voyage scale signature) で Industry-leading 1024-dim + Industry-leading 32K long context + Industry-leading Voyage scale signature。Industry-leading $0.06 cheaper (Industry-leading $0.06 USD per 1M tokens voyage-3 + $0.02 lite + Industry-leading cheaper than OpenAI $0.13 + Industry-leading Voyage pricing advantage) で Industry-leading $0.06 voyage-3 + Industry-leading $0.02 lite + Industry-leading cheaper than OpenAI + Industry-leading Voyage pricing。Industry-leading 3-tier models (Industry-leading voyage-3 flagship + voyage-3-lite cheaper + voyage-code-3 code-specialized + Industry-leading Voyage 3-tier signature + Industry-leading code-specialized voyage-code-3) で Industry-leading 3-tier + Industry-leading Voyage signature + Industry-leading code-specialized。Industry-leading Anthropic partnership (Industry-leading Anthropic recommended embedding partner + Industry-leading Voyage-Anthropic partnership signature + Industry-leading Claude embedding ecosystem) で Industry-leading Anthropic recommended + Industry-leading Voyage-Anthropic + Industry-leading Claude ecosystem。Industry-leading Tengyu Ma MIT alum (Industry-leading Tengyu Ma Stanford + MIT alum founders + Industry-leading Voyage AI 2023 founded + Industry-leading retrieval-optimized Voyage-3 cheaper OpenAI 2024) で Industry-leading Tengyu Ma Stanford + Industry-leading Voyage AI 2023 + Industry-leading retrieval-optimized 2024。但しIndustry-leading BGE + E5-Mistral + Nomic + OpenAI competition (Industry-leading BGE Large BAAI 1024-dim MIT 2023 + E5-Mistral Microsoft 4096-dim Mistral-7B 32K 2023 + Nomic Embed v1 768-dim Apache 2024 + OpenAI text-embedding-3-large 3072-dim Matryoshka $0.13/1M 2024 vs Voyage-3 1024-dim 32K retrieval-optimized $0.06/1M Anthropic partner 2024 trade-off) で Industry-leading 4-embedding model competitors vs Voyage-3 + Industry-leading retrieval-optimized + 1024-dim + 32K context + 3-tier voyage-3/lite/code-3 + $0.06/1M cheaper than OpenAI + Anthropic recommended partner + Tengyu Ma MIT alum Voyage AI 2024 unique advantage adoption alignment必須。
関連用語との違い
- vs OpenAI text-embedding-3-large (Closed $0.13 2024): Voyageは1024-dim + 32K + retrieval-optimized + $0.06 cheaper・OpenAIは3072-dim + 8191 + Matryoshka + $0.13
- vs BGE/E5-Mistral/Nomic (Open-source): VoyageはClosed API + retrieval-optimized + Anthropic partner・Open-sourceはMIT/Apache + weights public
- vs voyage-code-3: Voyage-3は general retrieval + 1024-dim・voyage-code-3はcode-specialized embedding
よくある質問（FAQ）
Q1: Voyage-3 vs OpenAI text-embedding-3-large 違いは？ A: Voyage-3 (Industry-leading retrieval-optimized specifically for RAG + 1024-dim default + 32K tokens max context + voyage-3 flagship + voyage-3-lite cheaper + voyage-code-3 code-specialized + top MTEB retrieval benchmarks + Anthropic recommended embedding partner + $0.06 USD per 1M tokens cheaper + Tengyu Ma MIT alum Voyage AI 2024) vs OpenAI text-embedding-3-large (Industry-leading 3072-dim default Matryoshka 256-3072 + 8191 tokens max context + MTEB 54.9% + multilingual 100+ languages + ChatGPT API ecosystem + $0.13 USD per 1M tokens + OpenAI 2024)・Industry-leading retrieval-optimized + 1024 + 32K + 3-tier + Anthropic partner + cheaper $0.06 = Voyage + Industry-leading 3072 + Matryoshka + 8191 + 100+ languages + ChatGPT + $0.13 = OpenAI preference judgment。
Q2: Industry-leading retrieval-optimized + 32K context value は？ A: Industry-leading retrieval + 32K (Industry-leading retrieval-optimized specifically for RAG + Industry-leading Voyage retrieval signature + Industry-leading top MTEB retrieval benchmarks 2024 + Industry-leading 32K tokens max context long context)。
Q3: Industry-leading $0.06 cheaper + Anthropic partnership value は？ A: Industry-leading $0.06 + Anthropic partnership (Industry-leading $0.06 USD per 1M tokens voyage-3 + Industry-leading $0.02 lite + Industry-leading cheaper than OpenAI $0.13 + Industry-leading Anthropic recommended embedding partner + Industry-leading Voyage-Anthropic partnership)。
まとめ
Voyage-3 = 2024年Voyage AI発表のretrieval-optimized embedding model。Industry-leading retrieval-optimized specifically for RAG + Industry-leading 1024-dim default + Industry-leading 32K tokens max context long context + Industry-leading voyage-3 flagship + voyage-3-lite cheaper + voyage-code-3 code-specialized 3-tier models + Industry-leading top MTEB retrieval benchmarks 2024 + Industry-leading multilingual support including code + Industry-leading $0.06 USD per 1M tokens voyage-3 + $0.02 lite pricing + Industry-leading cheaper than OpenAI $0.13 + Industry-leading API closed only + Industry-leading Anthropic recommended embedding partner + Industry-leading Voyage-Anthropic partnership + Industry-leading Tengyu Ma Stanford + MIT alum founders + Voyage AI 2023 founded + Industry-leading retrieval-optimized Voyage-3 cheaper than OpenAI 2024 position確立。