AI-Punkte Calculator

noris AI (nAI) is billed on a usage basis via AI-Punkte, token-based on a pay-as-you-grow principle. The token types and billing logic are explained in the concepts section.

The table shows the AI-Punkte per 1 million processed tokens. Enter your expected token volumes (in millions) per model to calculate your expected monthly consumption. The AI-Punkte catalog currently valid at noris governs.

Your inputs are never stored anywhere; they only live in your browser address bar. Use “Share input via link” below the table to share a filled-in calculation.

Model	Type	Tier	Number of tokens [1M]			AI-Punkte per 1M tokens			AI-Punkte/ month
			Input	Input (cached)	Output	Input	Cached	Output
GPT-OSS 120B	LLM (MoE)	LTS				6	1	25	0
Gemma 4 31B (IT)	Multimodal LLM	Productive				6	1	17	0
Qwen 3.6 27B	Multimodal LLM	Experimental				15	3	100	0
GLM 5.2	LLM (MoE)	Experimental				20	4	200	0
Harrier OSS v1 0.6B	Embedding	Experimental				1			0
BGE Reranker v2 M3	Reranker	Experimental				1			0
Total AI-Punkte/month									0

Request test access to the nAI

The button opens your email program with a pre-filled inquiry to noris. Your details only appear in this email draft; nothing is stored or sent automatically.

CompanyFirst name (contact person)Last name (contact person)Email (contact person)Street and house numberPostal code and cityOther

Open inquiry email

Share input via linkSave as PDF

As of July 16, 2026 (all information without guarantee)

Notes

• Reasoning tokens are billed at the same rate as output tokens, see What Is Reasoning?
• Cached input results from prefix caching and is billed at a reduced rate.
• Embedding and reranker models (Harrier, BGE) are billed exclusively via input tokens.
• Details on all models are available in the model overview.