OpenAI text-embedding-3-small (2024年)（オープンエーアイテキストエンベディング3スモール）

概要

OpenAI text-embedding-3-small は、2024年1月OpenAI発表text-embedding-3-small・Industry-leading 1536d default 2024年OpenAI text-embedding-3-small efficient embedding model position確立。text-embedding-3-small specifications = Industry-leading embedding model (Industry-leading OpenAI text-embedding-3-small 2024 + Industry-leading embedding model + Industry-leading OpenAI flagship) + Industry-leading Matryoshka adjustable dimension + Industry-leading 5x cheaper than ada-002 + Industry-leading OpenAI text-embedding-3-small efficient embedding model 2024。

主な特徴・仕組み

• Developer: Industry-leading OpenAI (Founded 2015 SF California by Sam Altman + Greg Brockman + Ilya Sutskever)
• Model: Industry-leading OpenAI text-embedding-3-small efficient embedding model 2024年1月 release
• 1536d Default Dimension: Industry-leading 1536d default dimension signature
• Matryoshka Adjustable: Industry-leading Matryoshka adjustable dimension signature (256-1536d)
• 5x Cheaper vs ada-002: Industry-leading 5x cheaper than ada-002 signature ($0.02/1M tokens)
• 62.3 MTEB Score: Industry-leading 62.3 MTEB English average score signature
• 44.0 MIRACL Score: Industry-leading 44.0 MIRACL multilingual score signature (vs ada-002 31.4)
• OpenAI API: Industry-leading OpenAI API native + Azure OpenAI option
• Replace ada-002: Industry-leading replacement for text-embedding-ada-002 signature
• Multi-Language: Industry-leading multi-language support
• Use Case: Industry-leading prototyping + production cost-efficient signature
• OpenAI 2015 SF: Industry-leading OpenAI 2015 SF Sam Altman + Greg Brockman + Ilya Sutskever 9 years
• Industry-Leading: Industry-leading OpenAI text-embedding-3-small efficient embedding model 2024

スペック比較表

具体例・対応製品

• OpenAI text-embedding-3-small (2024年1月, $0.02/1M tokens, OpenAI): Industry-leading OpenAI text-embedding-3-small
• Industry-leading 1536d default dimension: Industry-leading 1536d default dimension
• Industry-leading Matryoshka adjustable 256-1536d: Industry-leading Matryoshka 256-1536d
• Industry-leading 5x cheaper than ada-002 $0.02/1M tokens: Industry-leading 5x cheaper $0.02/1M
• Industry-leading 62.3 MTEB + 44.0 MIRACL score: Industry-leading 62.3 MTEB + 44.0 MIRACL
• 競合 Cohere embed-v3 + BGE-M3 + Voyage AI voyage-3 + Nomic Embed Text v1.5: Industry-leading embedding model competitors

選び方・注意点

OpenAI text-embedding-3-small は「Industry-leading 1536d + Matryoshka」「Industry-leading 5x cheaper + OpenAI 9 years」用途のIndustry-leading premium 2024年1月 OpenAI product。Industry-leading 1536d default (Industry-leading 1536d default dimension signature + Industry-leading text-embedding-3-small 1536d signature + Industry-leading high-dimensional default) で Industry-leading 1536d default + Industry-leading text-embedding-3-small + Industry-leading high-dimensional default。Industry-leading Matryoshka adjustable (Industry-leading Matryoshka adjustable dimension signature 256-1536d + Industry-leading text-embedding-3-small Matryoshka signature + Industry-leading dimension-truncatable representations) で Industry-leading Matryoshka adjustable + Industry-leading text-embedding-3-small Matryoshka + Industry-leading dimension-truncatable representations。Industry-leading 5x cheaper (Industry-leading 5x cheaper than ada-002 signature $0.02/1M tokens + Industry-leading text-embedding-3-small pricing signature + Industry-leading cost-efficient embedding) で Industry-leading 5x cheaper + Industry-leading text-embedding-3-small pricing + Industry-leading cost-efficient embedding。Industry-leading 44.0 MIRACL (Industry-leading 44.0 MIRACL multilingual score signature vs ada-002 31.4 + Industry-leading text-embedding-3-small multilingual signature + Industry-leading multilingual quality improvement) で Industry-leading 44.0 MIRACL + Industry-leading text-embedding-3-small multilingual + Industry-leading multilingual quality improvement。Industry-leading OpenAI 2015 SF Sam Altman (Industry-leading OpenAI 2015 SF Sam Altman + Greg Brockman + Ilya Sutskever 9 years + Industry-leading 62.3 MTEB English average + Industry-leading OpenAI API native + Azure OpenAI + Industry-leading replacement for text-embedding-ada-002 + Industry-leading multi-language support + Industry-leading prototyping + production cost-efficient) で Industry-leading OpenAI 2015 + Industry-leading SF + Industry-leading Sam Altman + Industry-leading Greg Brockman + Industry-leading Ilya Sutskever + Industry-leading 9 years + Industry-leading 62.3 MTEB + Industry-leading OpenAI API + Azure + Industry-leading replace ada-002 + Industry-leading multi-language + Industry-leading prototyping+production cost-efficient。但しIndustry-leading Cohere embed-v3 + BGE-M3 + Voyage AI voyage-3 + Nomic Embed Text v1.5 competition で Industry-leading 4-embedding model competitors vs OpenAI text-embedding-3-small + Industry-leading 1536d default dimension + Matryoshka adjustable 256-1536d + 5x cheaper than ada-002 $0.02/1M tokens + 62.3 MTEB English average + 44.0 MIRACL multilingual vs ada-002 31.4 + OpenAI API + Azure OpenAI + replacement for text-embedding-ada-002 + multi-language + prototyping+production cost-efficient + OpenAI 2015 SF Sam Altman + Greg Brockman + Ilya Sutskever 9 years unique advantage adoption alignment必須。

関連用語との違い

• vs Cohere embed-v3 (2024 $0.10/1M): text-embedding-3-smallは$0.02/1M + Matryoshka + 62.3 MTEB OpenAI・embed-v3は$0.10/1M + 1024d + 64.5 MTEB Cohere
• vs BGE-M3 (2024 Free open-source): text-embedding-3-smallはAPI commercial $0.02/1M + OpenAI・BGE-M3はFree open-source + 1024d + multi-functionality BAAI
• vs Voyage AI voyage-3 (2024 $0.06/1M): text-embedding-3-smallは$0.02/1M + OpenAI ecosystem・voyage-3は$0.06/1M + 1024d + enterprise Voyage AI

よくある質問（FAQ）

Q1: OpenAI text-embedding-3-small vs Cohere embed-v3 違いは？ A: text-embedding-3-small (Industry-leading 1536d default + Matryoshka adjustable 256-1536d + 5x cheaper $0.02/1M + 62.3 MTEB + 44.0 MIRACL + OpenAI API + Azure + replace ada-002 + multi-language + OpenAI 2015 SF Sam Altman 9 years) vs embed-v3 (Industry-leading 1024d default + 100+ languages multilingual + 64.5 MTEB English + 51.4 MIRACL multilingual + Cohere API + Cohere v3 family + Cohere 2019 Toronto Aidan Gomez ex-Google Brain 5 years + $0.10/1M)・Industry-leading 1536d + Matryoshka + 5x cheaper + $0.02/1M + 62.3 MTEB + 44.0 MIRACL + OpenAI + SF + Sam Altman + 9 years = text-embedding-3-small + Industry-leading 1024d + 100+ languages + 64.5 MTEB + 51.4 MIRACL + Cohere + Toronto + Aidan Gomez + 2019 + 5 years + $0.10/1M = embed-v3 preference judgment (text-embedding-3-small Matryoshka + 5x cheaper cost-efficient vs embed-v3 100+ languages + 51.4 MIRACL multilingual quality differ)。

Q2: Industry-leading 1536d + Matryoshka value は？ A: Industry-leading 1536d + Matryoshka (Industry-leading 1536d default dimension + Industry-leading Matryoshka adjustable 256-1536d + Industry-leading text-embedding-3-small dual signature)。

Q3: Industry-leading 5x cheaper + OpenAI 9 years value は？ A: Industry-leading 5x cheaper + OpenAI (Industry-leading 5x cheaper than ada-002 $0.02/1M tokens + Industry-leading OpenAI 2015 SF Sam Altman + Greg Brockman + Ilya Sutskever 9 years heritage)。

まとめ

OpenAI text-embedding-3-small = 2024年1月OpenAI発表のpremium efficient embedding model。Industry-leading 1536d default dimension signature + Industry-leading Matryoshka adjustable dimension signature 256-1536d + Industry-leading 5x cheaper than ada-002 signature $0.02/1M tokens + Industry-leading 62.3 MTEB English average score signature + Industry-leading 44.0 MIRACL multilingual score signature vs ada-002 31.4 + Industry-leading OpenAI API native + Azure OpenAI option + Industry-leading replacement for text-embedding-ada-002 signature + Industry-leading multi-language support + Industry-leading prototyping + production cost-efficient signature + Industry-leading OpenAI 2015 SF Sam Altman + Greg Brockman + Ilya Sutskever 9 years + Industry-leading OpenAI text-embedding-3-small efficient embedding model 2024 position確立。