Changelog

What changed on AI Tokens, newest first. A price increase, the removal of a model, or a change to the terms that limits your rights is announced by e-mail at least 30 days before it applies.

11 October 2026: AI Tokens opens#

• News on aitokens.ch/news, with an Atom feed, and by e-mail if you subscribe from the front page: we write first to confirm your address, and every e-mail has a link to unsubscribe.
• The dashboard's home shows this month's spend, requests and tokens, the last 30 days day by day, the models you use most and your latest requests, with the API address to copy. The menu is on the left.
• Card top-ups, from CHF 10 to CHF 5,000, on the Billing page of the dashboard, through Stripe, with more credit from CHF 50: +2% from CHF 50, +4% from CHF 100, +7% from CHF 500, +10% from CHF 1,000, +20% from CHF 5,000.
• Automatic top-up: when your credit falls below the threshold you set, we top up the amount you choose on your saved card, with 5% more credit. You switch it off when you like.
• In Switzerland: the models run on machines in Switzerland. Once a request reaches us, not a single packet of it leaves the Confederation.
• Sign-up is open. Confirm your e-mail address and CHF 2 of credit is loaded on your account, once. No card needed for it, no subscription.
• Prices in CHF, per model, input and output apart, excluding VAT: Pricing. Reasoning tokens are output tokens.
• Automatic model selection, "model": "auto": the model for each request, chosen in Switzerland; a task keeps its model through its tool calls; the headers say which model answered and why. See Automatic model selection.
• Spending limits per key, set in the dashboard: a request whose highest cost does not fit in what is left this month is refused with 402 spending_limit_reached before any model runs. See Errors.
• reasoning_effort and OpenRouter's reasoning object are accepted: none switches the reasoning of Qwen3.8 off; gpt-oss takes its three levels. See Every parameter.