Cohere API pricing per 1M tokens
Cohere API pricing starts at $0.0375 per 1M input tokens with Command R7B Arabic. The top tier is Command A Plus. Ruble figures use the CBR rate for Sep 18, 2026.
Prices in rubles as of 18.09.2026 at the CBR rate: 1 $ = 84.51 ₽.
Cohere API price per 1M tokens
| Model | Context | Input / 1M | Output / 1M | Capabilities | |
|---|---|---|---|---|---|
| North Mini Code 2026-06-09 | 256K | 0 ₽$0 | 0 ₽$0 | reasoningtoolsstructuredopen weights | |
| Command A Plus 2026-05-20 | 128K | 211 ₽$2.5 | 845 ₽$10 | reasoningtoolsvisionstructuredopen weights | |
| Command A Translate 2025-08-28 | 8K | 211 ₽$2.5 | 845 ₽$10 | toolsopen weights | |
| Command A Reasoning 2025-08-21 | 256K | 211 ₽$2.5 | 845 ₽$10 | reasoningtoolsopen weights | |
| Command A Vision 2025-07-31 | 128K | 211 ₽$2.5 | 845 ₽$10 | visionopen weights | |
| Command A 2025-03-13 | 256K | 211 ₽$2.5 | 845 ₽$10 | toolsopen weights | |
| Aya Vision 32B 2025-03-04 | 16K | — | — | visionopen weights | |
| Aya Vision 8B 2025-03-04 | 16K | — | — | visionopen weights | |
| Command R7B Arabic 2025-02-27 | 128K | 3,17 ₽$0.0375 | 12,68 ₽$0.15 | toolsopen weights | |
| Command R7B 2024-12-02 | 128K | 3,17 ₽$0.0375 | 12,68 ₽$0.15 | toolsopen weights | |
| Aya Expanse 32B 2024-10-24 | 128K | — | — | open weights | |
| Aya Expanse 8B 2024-10-24 | 8K | — | — | open weights | |
| Command R 2024-08-30 | 128K | 12,68 ₽$0.15 | 50,71 ₽$0.6 | toolsopen weights | |
| Command R+ 2024-08-30 | 128K | 211 ₽$2.5 | 845 ₽$10 | toolsopen weights |
Models shown: 14 / 14
About the Cohere API
Cohere is a Canadian company focused on enterprise workloads, with the Command A and Command R families (generation, document workflows, tool calling) and the North agent. Weights for part of the line-up are published, including the Aya Expanse and Aya Vision research models under CC-BY-NC-4.0.
Cohere publishes API prices per million tokens, billed separately for input and output, so the cost of a request depends on prompt and completion length. The table above lists 14 models with per-1M pricing, context window and maximum output.
Frequently asked questions
How much does the Cohere API cost?
There is no single rate — it depends on the model. The lowest input price across Cohere models is 3.17 ₽ per 1M tokens, converted from the provider's dollar price list, and the top-tier model is Command A Plus. The table above lists input and output rates for every Cohere model. Data as of September 2026.
Which Cohere model is the cheapest?
By input price it is Command R7B Arabic at 3.17 ₽ per 1M input tokens. The final bill also depends on the output rate, which varies far more between models: the longer the answers, the bigger the gap. Compare the Input / 1M and Output / 1M columns before you commit to a model.
How many Cohere models are in the catalog?
Cohere models in the catalog: 14, as of September 2026. The list is built automatically from the provider's public data — new models appear at the next sync, while retired ones stay in place with a deprecated status. Every model has its own page with prices and limits.
How are Cohere prices converted to rubles?
Cohere publishes its price list in dollars per 1M tokens, and the ruble figures are a conversion at the official Bank of Russia rate, currently 1 $ = 84.51 ₽. The rate and the date it was taken are shown in the page header, and the currency switch brings back the original dollar prices.
How do the Cohere model lines differ?
The lines differ in size, price and purpose: there are lightweight models starting at 3.17 ₽ per 1M input tokens, and top-tier ones such as Command A Plus. The table columns show exactly what changes — context window, input and output prices, capabilities. Details live on each model page.
How often are Cohere prices updated?
The catalog syncs every day at 06:00 Moscow time: provider price lists, limits, the model list and the Bank of Russia rate are all re-read. If Cohere changes a rate, the new price shows up here after the next sync. Current figures are as of September 2026.
Do Cohere models support reasoning?
Yes, some Cohere models support a reasoning mode — the Capabilities column flags it in the table, and each model page lists it separately. Keep in mind that reasoning consumes output tokens, so a request costs more even when the per-1M rate stays the same.
Does Cohere have open-weight models?
Yes, some Cohere models ship with published weights — the model page states this together with the license. You can run them on your own infrastructure, where the cost depends on your hardware rather than on a price list. Catalog prices refer to the provider's own API.
How much do 1,000 Cohere tokens cost?
Providers quote prices per 1M tokens, so 1,000 tokens cost exactly a thousandth of that. The cheapest option in the catalog is Command R7B Arabic at $0.0375 per 1M input tokens.