BroadRouter

River API pricing

Updated

River lists standard prices for 8 models. The cheapest is NVIDIA Nemotron 3.5 Super VL at USD 0.07 input and USD 0.37 output per 1M tokens; the dearest is Kimi K3 at USD 3 / USD 15. It publishes cached-input prices for all of them, and lists no per-model batch, flex or priority prices on the pages we read.

River hosts models it does not train. By maker: DeepSeek 2, Moonshot AI 2, Z.ai 2, NVIDIA 1, Qwen 1. 8 prices, each read from River's own pricing page, checked 2026-11-16. BroadRouter is BroadNet's own site and does not sell River's services.

Models with a standard price
8
Median input / output
USD 0.12 / USD 1.24
Cached-input prices
8 of 8
Cheapest host for
1 of 8 shared models

Which models is River the cheapest provider for?

Of the 8 models that River and at least one other provider both price at the standard tier, River has the lowest price for 1, ranked on a blend of 3 input tokens to 1 output token.

Models where River is the cheapest provider
ModelRiver input / outputNext cheapestSaving on 3:1 blendProviders
NVIDIA Nemotron 3.5 Super VLUSD 0.07 / USD 0.37Crusoe Cloud USD 0.10 / USD 0.4017%7

The 6 models River sells that the most other providers also sell, with River's rank among them on the 3:1 blend.

River on the most widely hosted models
ModelRiverRankCheapest
GLM 5.3USD 1.12 / USD 3.527 of 31DeepInfra USD 0.5625 / USD 2.50
GLM 5.3 FlashUSD 0.12 / USD 0.407 of 29GMI Cloud USD 0.075 / USD 0.25
DeepSeek V4.1 FlashUSD 0.12 / USD 0.482 of 23Relace USD 0.07 / USD 0.50
DeepSeek V4.1 Flash 1017USD 0.12 / USD 0.483 of 23Sail Research USD 0.09 / USD 0.30
Kimi K3USD 3 / USD 156 of 23Chutes USD 1.85 / USD 9.50
Kimi K3.1 ProUSD 2 / USD 104 of 20Sail Research USD 1.50 / USD 10

Which model makers does River sell?

River prices models from 5 makers. DeepSeek leads with 2 models. Standard prices, USD per 1M input / output tokens.

River models by maker
MakerModelsCheapest modelMedian input / output
DeepSeek2DeepSeek V4.1 Flash USD 0.12 / USD 0.48USD 0.12 / USD 0.48
Moonshot AI2Kimi K3.1 Pro USD 2 / USD 10USD 2.50 / USD 12.50
Z.ai2GLM 5.3 Flash USD 0.12 / USD 0.40USD 0.62 / USD 1.96
NVIDIA1NVIDIA Nemotron 3.5 Super VL USD 0.07 / USD 0.37—
Qwen1Qwen3.8 27B USD 0.10 / USD 2—

River price list

USD per 1M tokens: 8 prices for 8 models, one row per model with any other service tiers stacked under the standard price. Sources checked 2026-11-16 unless a row says otherwise. "vs maker" compares a standard price with the model maker's own API on the 3:1 blend; a minus means River is cheaper. Of 7 hosted model prices with a maker price to compare, 5 are below the maker's and 0 above.

River API prices by model
ModelTierInputCachedOutputContextvs makerSource
DeepSeek V4.1 FlashstandardUSD 0.12USD 0.0024USD 0.48not stated−60%source
DeepSeek V4.1 Flash 1017standardUSD 0.12USD 0.0024USD 0.48not stated−60%source
GLM 5.3standardUSD 1.12USD 0.208USD 3.52not stated−20%source
GLM 5.3 FlashstandardUSD 0.12USD 0.024USD 0.40not stated−20%source
Kimi K3standardUSD 3USD 0.30USD 15not statedsamesource
Kimi K3.1 ProstandardUSD 2USD 0.20USD 10not statedsamesource
NVIDIA Nemotron 3.5 Super VLstandardUSD 0.07USD 0.024USD 0.37not stated—source
Qwen3.8 27BstandardUSD 0.10USD 0.02USD 2not stated−49%source
  • Cached "—" means River's page shows no cache price for that row; budget cached tokens at the input price.
  • Context "not stated": River's pricing page does not give context windows.

Report a wrong price or number (email to info@broadnet.ai, with this page's link filled in).

What service tiers does River publish?

Standard prices only: River publishes no per-model batch, flex, priority or fast price on the pages we read.

Cached input: 8 of 8 standard prices include a cache-read price, a median 14% of the input price. What cached input means.

What we recorded about River's pricing page: Prices from the "Inference" table on river.ai/pricing (Prompt, Cached, Completion, USD per 1M tokens). The page fills this table with JavaScript from River's public pricing endpoint (console.river.ai/api/public/pricing, no login); values were read from that endpoint and the quotes are the rows as the page renders them. The page labels them "Preview rates, subject to change." Training and sampling prices on the same page are excluded. No context length is shown for inference.

Is River cheaper than the model makers' own APIs?

Mostly yes. Of 7 standard prices for models whose maker sells its own API, 5 are below the maker's price, 0 above and 2 the same, on the 3:1 blend. The median difference is 20% below the maker. The biggest saving is DeepSeek V4.1 Flash, 60% below DeepSeek.

BroadNet Inference and River

BroadRouter is BroadNet's own site, so BroadNet is never ranked on this page. BroadNet Inference also serves 3 models that River prices: DeepSeek V4.1 Flash at USD 0.175 / USD 0.52 against River's USD 0.12 / USD 0.48 (River is cheaper on the 3:1 blend); Qwen3.8 27B at USD 0.175 / USD 2.17 against River's USD 0.10 / USD 2 (River is cheaper on the 3:1 blend); GLM 5.3 at USD 0.78 / USD 3.46 against River's USD 1.12 / USD 3.52 (BroadNet is cheaper on the 3:1 blend). BroadNet prices are from its rate card, synced 2026-11-14. BroadNet Inference pricing.

More on River

Questions

How much does the River API cost per token?

River's standard prices run from USD 0.07 input and USD 0.37 output per 1M tokens (NVIDIA Nemotron 3.5 Super VL) to USD 3 / USD 15 (Kimi K3). The median across 8 standard prices is USD 0.12 input and USD 1.24 output per 1M, which is USD 0.00000012 per input token. Input, cached input and output are billed separately.

What is the cheapest model on River?

NVIDIA Nemotron 3.5 Super VL, at USD 0.07 input and USD 0.37 output per 1M tokens (USD 0.145 on a 3:1 input-to-output blend). The cheapest from other makers: DeepSeek V4.1 Flash (DeepSeek) at USD 0.12 / USD 0.48; Kimi K3.1 Pro (Moonshot AI) at USD 2 / USD 10; GLM 5.3 Flash (Z.ai) at USD 0.12 / USD 0.40.

Is River cheaper than other providers?

It depends on the model. Of 8 models that other providers also price, River is the cheapest standard price for 1, for example NVIDIA Nemotron 3.5 Super VL; another provider is cheaper for 7; and it is the dearest for none. Its median rank is 3.5 of a median 23 providers, on a 3:1 input-to-output blend.

Does River have batch, cached or priority pricing?

Cached input: yes, for 8 of 8 standard prices, at a median 14% of the input price. No batch, flex or priority prices are listed per model on the pages we read.

How much does a chatbot cost per month on River?

For 2,000 requests a day, 1,500 input tokens (60% cached), 300 output, over 30 days: USD 10.48 on NVIDIA Nemotron 3.5 Super VL, USD 13.09 on DeepSeek V4.1 Flash 1017 (the middle of River's list) and USD 394.20 on Kimi K3. Try your own numbers.

Read as MarkdownAsk ChatGPTAsk ClaudeAsk Perplexity