Models & pricing

Prices per million tokens, and where every model runs

Prices in USD per million tokens. You pay for the tokens you use, from a prepaid balance. Launch pricing, until November 30, 2026.

Same values as the X-Data-Location header.

OpenRouter column: the minimum price listed across all providers of each model, refreshed every day at 00:00 UTC. Last update: Oct 6, 2026, 00:00 UTC. Margin: what you keep if you buy from us and resell at the OpenRouter minimum, as a share of that price. Our prices and OpenRouter's can both change from one day to the next, and so can the differences shown.

■ margin 30 % or more
■ margin 10–30 %
■ margin 0–10 %
■ negative margin

Prices in USD per million tokens and hosting region for each model
Model	Input	Cached input	Output	OpenRouter (min.)	Margin vs OpenRouter	Context	Region	Status
Apertus v1.5 70B Swiss model`swiss-ai/apertus-v1.5-70b`	$0.875	$0.19	$3.125	Not on OpenRouter	—	64k	CHSwitzerland	available
gpt-oss-120b`openai/gpt-oss-120b`	$0.10	$0.10	$0.50	$0.03 in · $0.17 outmin. of 23 providers	−194 % output−233 % input	128k	EUEuropean Union	available
gpt-oss-20b`openai/gpt-oss-20b`	$0.015	$0.008	$0.075	$0.018 in · $0.09 outmin. of 12 providers	+17 % output+17 % input	64k	CHSwitzerland	available
Kimi K2.6`moonshotai/kimi-k2.6`	$0.85	$0.85	$4.15	$0.465 in · $2.40 outmin. of 18 providers	−73 % output−83 % input	128k	CHSwitzerland	available
Qwen3.8-27B`qwen/qwen3.8-27b`	$0.10	$0.02	$1.25	$0.024 in · $1.49 outmin. of 18 providers	+16 % output−317 % input	128k	EUEuropean Union	available
Apertus v1.5 8B Swiss model`swiss-ai/apertus-v1.5-8b`	$0.10	$0.03	$0.30	Not on OpenRouter	—	64k	CHSwitzerland	best effortnot running now
Qwen3.6-35B-A3B`qwen/qwen3.6-35b-a3b`	—	—	—	—	—	32k	EUEuropean Union	committedDetails

Committed throughput

Some models are not sold per token but with a throughput commitment on our API: a guaranteed output rate for your account, prepaid by the week. Tokens are included up to the guaranteed throughput.

Qwen3.6-35B-A3B

qwen/qwen3.6-35b-a3b

EU · European Union

1,800 output tokens per second guaranteed, up to 64 concurrent requests (28.1 tokens per second per request), measured every hour. $4.95 per hour ($831.60 per week), prepaid, 1-week minimum. Active within 72 hours of payment. Served from SOKKAN infrastructure in the EU zone (European Union).

Request committed throughput

Committed throughput is a service level of our API: NINABOT chooses the compute that serves it within the zone shown, and throughput you do not use may serve other traffic while your guarantee is met. Not for sensitive personal data (official identity numbers, financial account data, payment card data, health data). Measurement rules, service credits and renewal are in the Terms of Service. Requests are handled by email for now: the button opens a message to support, and we reply with a prepaid invoice. Models marked “best effort” are served per token, without any guarantee, only while their capacity is running.

Notes

Cached input

Prompt tokens read from the inference cache. Where cache hits are not reported, the whole prompt is billed at the input price.

Region

Where the model runs: CH (Switzerland), EU (European Union / EEA) or US. Every response repeats it in x-sokkan-region.

Coming soon

Listed with the region where we plan to serve it. Calls return 503 until the model is live.

Catalogue as JSON

The same catalogue, with prices and regions, is at GET /v1/models.

Need steady volume or a model that is not listed? Write to [email protected].