Rerank cost calculator
Hosted rerankers do not bill the same way. Cohere Rerank 4 is priced per search — one query plus up to 100 documents, with any document over 500 tokens counted as several. Google and Amazon Bedrock charge per query of up to 100 documents, and Pinecone per request of up to 100, whatever their length. Cohere Rerank 5, Voyage, Jina, ZeroEntropy, Contextual AI and Alibaba Cloud charge per token — Jina counting one request's tokens, the query plus every candidate — and so do the yuan-priced services in mainland China. Which is cheapest depends on how your passages round. Put your own numbers in.
Your workload
Monthly cost
| Option | Billing unit | Billable volume | Monthly cost |
|---|
Priced in yuan, for mainland China
| Option | Billing unit | Billable volume | Monthly cost |
|---|
Why the ranking changes
Cohere Rerank 4 and Voyage both end up charging for text, but they round it differently. Cohere bills each candidate as one document per started 500 tokens — query included — and each search covers 100 documents. Voyage bills exactly the tokens you send, counting the query once per document.
At list prices the two meet exactly: 500 tokens costs the same on Cohere Rerank 4 Pro ($0.0025 per search of 100 documents) and Voyage rerank-3 ($0.05 per 1M tokens) — Voyage’s own pricing page uses that very example. What separates them is rounding:
- Short passages — Cohere still bills each one as a full document, so per-token billing wins, by more the shorter they are.
- Just past a 500-token boundary — a 520-token candidate is two documents on Cohere; per-token billing costs about half.
- Top-k that isn’t a multiple of 100 — Cohere rounds each query up to a whole search, so cost steps up rather than sliding.
- Cohere Rerank 4 Fast vs Voyage rerank-3 — Fast is cheaper only when query plus passage fills more than 80% of its last chunk. The
-litetier is cheaper than both on any workload.
Rerank 5 removes the rounding: Cohere bill it per token, at the same list prices as Voyage — $0.05 per 1M tokens for Pro against $0.05 for rerank-3, and $0.02 for Fast against $0.02 for rerank-3-lite. Cohere don’t say how the query is counted, so the calculator counts it as Voyage does, and the two pairs come out level. Choose between them on quality, not price.
Jina charges the same $0.05 per 1M tokens, but counts them differently again: v3.5 is listwise, so the query and every candidate share one 131K-token context and the query is paid for once per query, not once per document. On a long candidate list that is a small saving — and it is not why you would pick Jina. What it buys you is the listwise comparison itself; the tokens are the cheaper part of the story.
Correction, October 2026: until now this page said Cohere’s 500-token splitting was a myth and that per-search pricing ignores passage length. Cohere’s own pricing FAQ says otherwise: a document over 500 tokens, counting the query, is split, and every chunk counts toward the 100 documents in a search. The calculator now models that, which removes the long-passage advantage the old version showed for Cohere.
Your workload: —
What about self-hosting?
Self-hosting has no per-call price, so it does not belong in the table above — you trade a usage bill for a GPU bill plus operations. The comparison only becomes meaningful at your own volume: divide your monthly GPU cost by the number of queries above and compare that to the per-query figures in the table. Below a few hundred thousand queries a month an API is usually cheaper than a dedicated GPU; well above that, self-hosting a Qwen3-Reranker or bge-reranker-v2-m3 starts to pay off. Our self-hosting guide covers the serving side.
Note that a smaller model is not automatically a worse one — Jina Reranker v3.5 reaches 63.20* BEIR nDCG@10 at 0.6B in Jina's own test, level with a 4B model, so the cheapest thing to serve may also be among the most accurate.
Assumptions and sources
- Cohere Rerank 4 — $0.0025 per search (Pro) and $0.002 per search (Fast). A search is one query plus up to 100 documents; per Cohere’s pricing FAQ, a document longer than 500 tokens including the query is split, and each chunk counts as a document. These are the last prices Cohere listed ($2.50 and $2 per 1,000 searches): its pricing page stopped showing Rerank 4 on 21 October 2026, when Rerank 5 was priced, though Rerank 4 is still served.
- Voyage rerank-3 — $0.05 per 1M tokens, and $0.02 for
rerank-3-lite. Billable tokens are the query counted once per document, plus every document token you send — so 50 candidates means the query is billed 50 times. The olderrerank-2.5and-litecost the same per token. Free tokens and the Batch API discount are not applied here — see the Voyage page for which models they cover. - Jina Reranker v3.5 — $0.05 per 1M tokens, from Jina’s model-discovery API (api.jina.ai/v1/models), which lists the rate for
jina-ai/jina-reranker-v3.5as a price per token; jina.ai’s own pages quote no figure and sell prepaid token packages instead. Jina describe the Reranker API’s usage as “the number of tokens in the input request”, and v3.5 takes the query and all candidates in one context window, so this page counts the query once per request rather than once per document. The free trial tokens a new key comes with are not applied. - Google Vertex AI ranking — $1.00 per 1,000 queries; a query covers up to 100 documents, and each further 100 counts as another query (Google’s pricing). Passage length doesn’t change the price, but the models read only 1,024 tokens per record.
- Alibaba Cloud Model Studio — qwen3-rerank at $0.1 per 1M input tokens internationally and ¥0.5 in the Beijing region, where qwen3.7-text-rerank costs the same (price list). Alibaba bills input tokens and caps a request at query tokens × documents + document tokens, so this page counts tokens the way it does for Voyage. qwen3-rerank rejects any single document over 4,000 tokens rather than truncating it.
- SiliconFlow — bge-reranker-v2-m3 is listed free and its Pro tier at ¥0.07 per 1M tokens (price list). SiliconFlow doesn’t say how it counts rerank tokens; this page assumes the same formula as Alibaba and Voyage. Its price list gives no price for the Qwen3-Reranker models it also serves.
- ZeroEntropy zerank-2 — $0.025 per 1M tokens (ZeroEntropy’s docs). The docs don’t say how a rerank request’s tokens are counted; this page counts them like Voyage’s, the query once per document, which is also how ZeroEntropy counts a request against its rate limits.
- Contextual AI Rerank API — $0.05 per 1M tokens for v2 and $0.02 for v2-mini (Contextual pricing). As with ZeroEntropy, the query is counted once per document; Contextual don’t say how they count it.
- Amazon Bedrock, Amazon Rerank 1.0 — $1.00 per 1,000 queries; a query covers up to 100 documents, and more count as further queries (Bedrock pricing). Cohere Rerank 3.5 on Bedrock is left out: the same page says a document holds up to 500 tokens and that one over 512 is split, and the two figures don’t agree.
- Pinecone — bge-reranker-v2-m3 and the deprecated pinecone-rerank-v0 at $2 per 1,000 requests on the paid plans; a request takes up to 100 documents, so more need further requests (Pinecone pricing). The included free requests aren’t applied, and cohere-rerank-4-fast is left out because Pinecone doesn’t define the “rerank unit” it bills it in.
- Cohere Rerank 5 — $0.05 per 1M tokens (Pro) and $0.02 (Fast), from cohere.com/pricing. Cohere don’t say how a rerank request’s tokens are counted; this page counts them like Voyage’s, the query once per document.
- Token counts are approximate — every vendor tokenises differently, and this page multiplies your averages rather than tokenising real text.
- Rates verified in October 2026 and re-checked daily against each vendor’s price page by this site’s source check. Confirm before committing to a budget.
Cost is only half the decision
Quality is the other half. Run a cross-encoder against your own passages in the browser — no key, no spend.
Open the live demo →