OpenAI text-embedding-3-large
The battle-tested managed embedding API with shrinkable 3072-dim vectors
About
text-embedding-3-large is OpenAI's flagship embedding model, released January 25, 2024 alongside the small variant. It outputs up to 3072-dimensional vectors over an 8191-token window, with a dimensions parameter (Matryoshka training) that lets you request shorter vectors — dropping to 256 dims still beats the older ada-002 at full size, per OpenAI's reported figures. It scores a reported 64.6 on MTEB and 54.9 on MIRACL, and costs $0.13 per million input tokens ($0.065 on the Batch API). It remains the default choice for teams already on OpenAI's platform because of ecosystem fit and operational simplicity. The tradeoffs: it is API-only, 8K context is half of what newer rivals offer, and independent comparisons now rank several competitors ahead of it on retrieval quality.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
MTEB average 64.6; MIRACL 54.9 (OpenAI-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: January 25, 2024 release date, 3072 dims, 8191 tokens, $0.13/1M pricing, MTEB 64.6, MIRACL 54.9 all verified in a March 2026 model card and OpenAI's launch coverage read this session.; Batch API $0.065 rate noted separately from standard pricing.; Independent comparisons ranking newer rivals ahead are disclosed honestly..