one-api pricing: LLM tokens and lookups

How the price is set

The Cloudflare list row is the provider price on Cloudflare's pricing pages, checked 2026-10-22. Unified Billing adds 5% when prepaid credits are bought, so the cash floor is the list price divided by 0.95: what a token costs us before any margin. The one-api.ai row adds the posted margin: 4% for DeepSeek V4.1 Flash and DeepSeek V4 Pro, 15% for Claude Haiku 5.5 and Grok 4.7, and 50% for GLM 5.3, rounded to the cent for GLM and to a tenth of a cent for the rest.

Calls are billed in credits, tracked to a millionth of a credit. A call first reserves its worst case, worked out from the prompt size and max_tokens, and refuses with HTTP 402 if the balance cannot cover it. After the provider answers, the call settles to the reported usage at exactly the posted per-token rate, with no rounding up per call. The model pages show worked examples.

Credit packs

When billing opens, credits are sold in packs of $10.00 and $50.00, paid once, with no subscription. Credits do not expire while the service runs; unused credits can be refunded under the refund policy and cannot be withdrawn as cash or moved to another account.

DeepSeek V4.1 Flash

Model id @cf/deepseek-ai/deepseek-v4.1-flash-1017. Part of early access. Cloudflare caps it at 50 requests per minute when paid with prepaid credits. hiapi posts $0.44 input and $1.32 output per 1M tokens (collected 2026-10-21).

Per 1M tokens

Source: Cloudflare Workers AI pricing, page updated 2026-10-20, checked 2026-10-22.

DeepSeek V4 Pro

Model id @cf/deepseek-ai/deepseek-v4-pro-0813. Part of early access. Cloudflare caps it at 50 requests per minute when paid with prepaid credits. hiapi posts $1.50 input and $4.50 output per 1M tokens (collected 2026-10-21).

Per 1M tokens

Source: Cloudflare Workers AI pricing, page updated 2026-10-20, checked 2026-10-22.

GLM 5.3

Model id @cf/zai-org/glm-5.3. Part of early access. Cloudflare caps it at 50 requests per minute when paid with prepaid credits. hiapi posts $4.40 input and $15.40 output per 1M tokens (collected 2026-10-21).

Per 1M tokens

Source: Cloudflare Workers AI pricing, page updated 2026-10-20, checked 2026-10-22.

Claude Haiku 5.5

Model id anthropic/claude-haiku-5.5. Published for comparison; not part of early access. hiapi does not post a price for it.

Prompts up to 100,000 tokens, per 1M

Prompts over 100,000 tokens, per 1M

Source: Cloudflare model page, checked 2026-10-22.

Grok 4.7

Model id xai/grok-4.7. Published for comparison; not part of early access. hiapi does not post a price for it.

Under 200k input tokens, per 1M

At or above 200k input tokens, per 1M

Source: Cloudflare model page, checked 2026-10-22.

Lookups

Each lookup costs one credit per call. Every account gets 25 free lookups per UTC month, and a lookup that fails with HTTP 400 or a 5xx error is released rather than charged.