Text embeddings

Qwen3-Embedding-4B

Alibaba Qwen · open weights

$0.020 per 1M input tokens, cheapest of 2 sellers (DeepInfra)

Dominated: Qwen3-Embedding-8B scores higher and costs less. As of .

What was measured

BoardMean (Task)Mean (TaskType)TasksRevision
MTEB(eng, v2)74.6168.141 of 41636cd9bf47d976946cdbb2b0c3ca0cb2f8eea5ff

By task type: Classification 89.84 · Clustering 57.51 · PairClassification 87.01 · Reranking 50.76 · Retrieval 68.46 · STS 88.72 · Summarization 34.39

What sellers charge

Spread across 2 distinct sellers: between cheapest and dearest at the reference unit.

SellerAs the seller states itConditionsNormalised per 1M input tokensSource
DeepInfra2e-8 per_token$0.020 per token × 1,000,000api.deepinfra.com read 2026-09-19 · feed
OpenRouter2e-8 per_tokencontext 32768$0.020 per token × 1,000,000openrouter.ai read 2026-09-19 · feed

Provenance

MTEB results, https://github.com/embeddings-benchmark/results, released under CC0-1.0; MTEB(eng, v2) means computed from the per-task files as described in data/modalities/embeddings/benchmark.json.