EmbedPay vs OpenAI Embeddings (direct)

EmbedPay is an x402 wrapper around the same OpenAI text-embedding-3-small model. Direct OpenAI is cheaper per token but requires an account, API key, and rate-limit management. EmbedPay's premium buys signup-free access with composable x402 billing.

At a glance

EmbedPay OpenAI Embeddings (direct)
Underlying modeltext-embedding-3-small (1536d)Same — text-embedding-3-small (1536d)
Signup requiredNoYes — OpenAI account + billing card + API key
Per 1k tokens (standard)$0.00005 USDC$0.00002 USD
Per 1k tokens (batch ≥100 inputs)$0.00003 USDCn/a — batch API has separate quota
Premium vs direct~2.5× direct rate. The premium covers x402 infrastructure + no signup convenience.
Token counting transparencycl100k_base counted server-side, returned in input_tokens field — you can verify locallySame tokenizer, count returned in usage block
Rate limits600 req/min per IP · 10M tokens/IP/hourAccount-tier based (3k–10k+ RPM)
Model version pinningReturned in response: model_version: "text-embedding-3-small (2024-01-25)"Implicit — OpenAI rev-locks behind the scenes
MCP integrationNative via @melis-ai/x402-tools-mcpVia OpenAI SDK wrappers
Multi-tenancyPer-agent pay-per-call — no shared accountOne account = one billing relationship

Cost math at three volumes

VolumeEmbedPayOpenAI directPremium
1M tokens$0.05$0.02+$0.03
100M tokens$5.00$2.00+$3.00
1B tokens$50.00$20.00+$30.00
1B tokens, batch tier$30.00$20.00+$10.00

At ~1B tokens you're $30 better off direct — and at that scale you probably want the dashboard, rate-limit insights, and support contract that come with an OpenAI org. EmbedPay's sweet spot is below ~100M tokens/month where the operational overhead of provisioning + securing an OpenAI key outweighs the price delta.

Choose EmbedPay when

  • You're building an agent that needs embeddings but you don't want to manage an OpenAI key.
  • Multi-tenant agent runtime where customers shouldn't share an OpenAI billing relationship.
  • You're already in the x402 ecosystem — same wallet pays for ScrapePay → MarkdownOpt → EmbedPay → MemoryServe.
  • Total volume under ~100M tokens/month and the wrapper premium is < engineering time saved.
  • Need honest token-counting in the response for billing reconciliation across agents.

Choose OpenAI direct when

  • High and stable volume — 1B+ tokens/month makes the per-token delta material.
  • You already have an OpenAI account with negotiated enterprise rates.
  • You need OpenAI's other embedding models (text-embedding-3-large at 3072d, ada-002) and want a single provider for all of them.
  • You want dashboard analytics, fine-grained rate-limit visibility, and a support contract.

Try EmbedPay

Install once: npx @melis-ai/x402-tools-mcp. Then call the embedpay tool from your agent. See the EmbedPay service page for the full schema and the canonical RAG pipeline for a worked composition example.