LLM API provider price database
Updated
1,671 LLM API prices for 430 models from 41 providers, one row per price, each read from the provider's own pricing page. Every row has input, cached-input and output price per 1M tokens, tier, context, the source URL, the date checked (2026-11-14 to 2026-11-15) and a verbatim quote. Download as CSV or JSON under CC BY 4.0.
Download CSV Download JSON Croissant metadata Compilation CC BY 4.0 · prices credited to each provider · checked 2026-11-14 to 2026-11-15
BroadRouter is BroadNet's own site. This database holds third-party prices only, read from each provider's own page; BroadNet's prices are in its pricing.json. Per-model statistics built from these rows are in the LLM API price index.
Key findings
- 1,671 prices from 41 providers' own pricing pages, covering 430 models. Every row has a source URL, a check date and a verbatim quote (1,671 of 1,671 rows quoted), from 363 distinct source pages.
- 1,218 rows are standard on-demand prices; the rest are other service tiers: 157 batch, 117 flex / spot, 106 priority, 70 fast, 3 ultrafast.
- 198 models have a standard price from two or more providers and 155 from three or more. 230 have a single provider. GLM 5.3 has the most: 24 providers.
- 328 rows are a model maker selling its own model on its own API (16 of the 41 providers are makers).
- 65.2% of standard rows include a cached-input price. 47.6% of rows state a context window, 71.3% give the provider's model ID, and only 24.1% state the numeric precision served.
- Check dates run from 2026-11-14 to 2026-11-15. A weekly script re-fetches every source page; flagged rows are re-read and updated.
Which providers are included?
41 providers, by number of prices. Each provider's rows come from the pricing page linked here and the pages it links to.
| Provider | Kind | Prices | Models | Tiers | Checked |
|---|---|---|---|---|---|
| DeepInfra | host | 183 | 92 | standard, flex, priority | 2026-11-14 |
| Amazon Bedrock | host | 146 | 110 | standard, batch, flex, priority, fast | 2026-11-15 |
| Microsoft Azure | host | 124 | 97 | standard, batch, flex, priority, fast | 2026-11-14 |
| Google Vertex AI | host | 118 | 63 | standard, batch, priority | 2026-11-14 |
| OpenAI | maker | 107 | 37 | standard, batch, flex, fast, ultrafast | 2026-11-14 |
| GMI Cloud | host | 97 | 97 | standard | 2026-11-15 |
| DigitalOcean | host | 95 | 59 | standard, flex, fast | 2026-11-14 |
| Featherless | host | 87 | 87 | standard | 2026-11-15 |
| Novita AI | host | 68 | 67 | standard, fast | 2026-11-14 |
| Venice | host | 60 | 57 | standard, fast, ultrafast | 2026-11-15 |
| SiliconFlow | host | 56 | 56 | standard | 2026-11-15 |
| Google Gemini API | maker | 48 | 12 | standard, batch, flex, priority | 2026-11-14 |
| Alibaba Cloud | maker | 44 | 43 | standard, fast | 2026-11-14 |
| Anthropic | maker | 41 | 18 | standard, batch, fast | 2026-11-14 |
| Cloudflare Workers AI | host | 40 | 40 | standard | 2026-11-14 |
| Parasail | host | 40 | 39 | standard, fast | 2026-11-14 |
| Fireworks AI | host | 36 | 15 | standard, flex, priority, fast | 2026-11-14 |
| Mistral AI | maker | 27 | 9 | standard, batch, priority | 2026-11-15 |
| Amazon (Bedrock) | maker | 25 | 7 | standard, batch, flex, priority, fast | 2026-11-15 |
| Together AI | host | 25 | 25 | standard | 2026-11-14 |
| Perplexity | maker | 24 | 24 | standard | 2026-11-15 |
| Z.ai | maker | 21 | 21 | standard | 2026-11-14 |
| CoreWeave Forge | host | 20 | 20 | standard | 2026-11-15 |
| Chutes | host | 19 | 18 | standard, fast | 2026-11-15 |
| Baseten | host | 18 | 14 | standard, fast | 2026-11-14 |
| Baidu | maker | 15 | 14 | standard, fast | 2026-11-15 |
| Hyperbolic | host | 11 | 11 | standard | 2026-11-14 |
| Moonshot AI | maker | 10 | 4 | standard, batch, fast | 2026-11-14 |
| Crusoe Cloud | host | 9 | 9 | standard | 2026-11-15 |
| IBM watsonx.ai | maker | 9 | 9 | standard | 2026-11-15 |
| xAI | maker | 8 | 8 | standard | 2026-11-14 |
| AkashML | host | 7 | 7 | standard | 2026-11-15 |
| Cohere | maker | 6 | 6 | standard | 2026-11-14 |
| FriendliAI | host | 5 | 5 | standard | 2026-11-15 |
| SambaNova | host | 5 | 5 | standard | 2026-11-14 |
| Groq | host | 4 | 4 | standard | 2026-11-14 |
| Nscale | host | 4 | 4 | standard | 2026-11-15 |
| MiniMax | maker | 3 | 2 | standard, priority | 2026-11-14 |
| AI21 Labs | maker | 2 | 2 | standard | 2026-11-15 |
| Cerebras | host | 2 | 2 | standard | 2026-11-14 |
| DeepSeek | maker | 2 | 2 | standard | 2026-11-14 |
Method
- Source: each provider's own pricing page or price documentation, read with AI assistance and a page renderer between 2026-11-14 and 2026-11-15. Every row stores the source URL, the date checked and a verbatim quote from the page. No router, reseller or aggregator feed is used.
- Scope: 41 providers (16 of them model makers with their own API), text and chat models only. Image, audio and embedding models are out of scope. Where a provider shows no public per-token price (for example "contact sales"), the model is left out and the provider file's notes say why.
- Prices are USD per 1M tokens as the page shows them, at the lowest prompt-size tier in the provider's default region. Batch, flex or spot, priority, fast and ultrafast prices are separate rows with their tier. Promotional prices are kept as published, with the condition in notes. DeepSeek rows carry the peak-hour price; the off-peak price is in notes.
- Model slugs join rows for the same model across providers. A dated snapshot that a maker serves under a base name keeps its own slug.
- Re-checks: every Monday a script re-fetches each source URL and flags rows whose quote or price it can no longer find. It never edits data; flagged pages are re-read with AI assistance, and the value, quote and date are updated. Each row keeps its own check date.
- BroadNet Inference, which publishes this site, is not in the database; its prices are in its own machine-readable file, pricing.json.
Limits: published list prices, not what any customer paid. One region and the lowest prompt-size tier per provider. Prices are read with AI assistance and a page renderer, so a row can be wrong; the quote lets you check it, and our collection rules say how errors are fixed. Code: src/lib/dataset-provider-prices.ts.
What do the columns mean?
| Column | Meaning |
|---|---|
provider | Provider ID (one source file per provider: src/data/own-prices/<provider>.json). |
provider_name | Provider's product name, as the file records it. |
provider_kind | "maker" if the provider trains models and sells them on its own API, else "host". |
maker_api | true when this row is the model's maker selling it on its own API (a maker that hosts another lab's model is a host for that row). |
model | Model slug, as in the /llm-api-pricing/<model>/ page address. Rows with the same slug are the same model. |
name | Model name as this provider lists it. |
maker | Lab that trained the model. |
provider_model_id | The provider's own model ID for API calls, where its page or docs give one; empty if not stated. |
tier | Service tier: standard (standard), batch (batch), flex (flex / spot), priority (priority), fast (fast), ultrafast (ultrafast). Only standard rows are ranked on the site. |
input_per_m | Input price, USD per 1M tokens, at the lowest prompt-size tier in the default region. |
cached_input_per_m | Cache-read input price, USD per 1M tokens; empty if the page shows none. |
output_per_m | Output price, USD per 1M tokens (reasoning tokens bill as output). |
context | Context window in tokens as this provider states it; empty if not stated. |
precision | Numeric precision served (for example fp8), only where the provider states it; empty means not stated, not full precision. |
source_url | The provider page the price was read from. |
checked | Date the price was read from that page (YYYY-MM-DD). |
quote | Verbatim text from the source page that shows the price, so anyone can re-check it. |
notes | Region, promotion, peak or off-peak, batch or other conditions that apply to this row. |
The CSV and JSON hold the same rows. The JSON adds the method, this column list, the licence and a per-provider summary.
How to cite this dataset
Our compilation (the selection, normalised fields, model matching, notes, tables and method): CC BY 4.0. Free to use, share and adapt with attribution and a link to this page.
Each price is a fact published by the provider named in its row, at the source URL given per row, and each quote is a short excerpt from that page; we do not license those pages to you. Credit the provider as the source of each price.
Plain text
BroadRouter (2026). LLM API provider price database: every price with its source, date and quote (version 2026-11-15) [Data set]. BroadNet Technologies. https://broadrouter.com/data/llm-api-provider-prices/BibTeX
@misc{broadrouter_llm_provider_prices,
author = {BroadRouter},
title = {LLM API provider price database: every price with its source, date and quote},
year = {2026},
version = {2026-11-15},
publisher = {BroadNet Technologies},
howpublished = {\url{https://broadrouter.com/data/llm-api-provider-prices/}},
note = {Data set, data as of 2026-11-15},
}Where is this data used?
- LLM API pricing by model: every provider's price for each model, ranked.
- The cheapest LLM APIs, ranked on one blended price.
- LLM API price index: medians, spreads and maker-versus-host figures per model.
- Price change log: what moved between re-checks.
- LLM API cost calculator: monthly cost on any of these models.
Questions
What is in the LLM API provider price database?
One row per price: 1,671 prices from 41 providers for 430 text models. Each row gives the provider, model, the provider's model ID, service tier, input, cached-input and output price per 1M tokens, context, precision, the source URL, the date checked and a quote from the page.
How is it different from the LLM API price index?
The price index is the analysis: one row per model with medians, spreads and maker-versus-host figures. This database is the raw input to it and to every model price page: each individual price, with the evidence to check it.
How are the prices collected and kept current?
Each provider's own pricing page is read, with AI assistance or a page renderer for pages that load prices with JavaScript, and stores the price with the source URL, the date and a verbatim quote. No aggregator feed is used. Every Monday a script re-fetches the source pages and flags rows it can no longer match; those pages are re-read and the rows updated. Changes appear in the price change log.
Can I use this data?
Yes. Our compilation (the selection, normalised fields, model matching, notes and method) is CC BY 4.0: cite it as shown under "How to cite". Each price is a fact published by the provider named in its row; credit that provider as the source of the price.
Why is BroadNet not in it?
BroadRouter is BroadNet's own site, so the database holds third-party prices only. BroadNet Inference publishes its prices in a machine-readable file, pricing.json, and on the pricing page.
Read as MarkdownAsk ChatGPTAsk ClaudeAsk Perplexity
Sources
- DeepInfra: pricing (checked 2026-11-14)
- Amazon Bedrock: pricing (checked 2026-11-15)
- Microsoft Azure: pricing (checked 2026-11-14)
- Google Vertex AI: pricing (checked 2026-11-14)
- OpenAI: pricing (checked 2026-11-14)
- GMI Cloud: pricing (checked 2026-11-15)
- DigitalOcean: pricing (checked 2026-11-14)
- Featherless: pricing (checked 2026-11-15)
- Novita AI: pricing (checked 2026-11-14)
- Venice: pricing (checked 2026-11-15)
- SiliconFlow: pricing (checked 2026-11-15)
- Google Gemini API: pricing (checked 2026-11-14)
- Alibaba Cloud: pricing (checked 2026-11-14)
- Anthropic: pricing (checked 2026-11-14)
- Cloudflare Workers AI: pricing (checked 2026-11-14)
- Parasail: pricing (checked 2026-11-14)
- Fireworks AI: pricing (checked 2026-11-14)
- Mistral AI: pricing (checked 2026-11-15)
- Amazon (Bedrock): pricing (checked 2026-11-15)
- Together AI: pricing (checked 2026-11-14)
- Perplexity: pricing (checked 2026-11-15)
- Z.ai: pricing (checked 2026-11-14)
- CoreWeave Forge: pricing (checked 2026-11-15)
- Chutes: pricing (checked 2026-11-15)
- Baseten: pricing (checked 2026-11-14)
- Baidu: pricing (checked 2026-11-15)
- Hyperbolic: pricing (checked 2026-11-14)
- Moonshot AI: pricing (checked 2026-11-14)
- Crusoe Cloud: pricing (checked 2026-11-15)
- IBM watsonx.ai: pricing (checked 2026-11-15)
- xAI: pricing (checked 2026-11-14)
- AkashML: pricing (checked 2026-11-15)
- Cohere: pricing (checked 2026-11-14)
- FriendliAI: pricing (checked 2026-11-15)
- SambaNova: pricing (checked 2026-11-14)
- Groq: pricing (checked 2026-11-14)
- Nscale: pricing (checked 2026-11-15)
- MiniMax: pricing (checked 2026-11-14)
- AI21 Labs: pricing (checked 2026-11-15)
- Cerebras: pricing (checked 2026-11-14)
- DeepSeek: pricing (checked 2026-11-14)