Touchmark operates a live order book for reserved AI model-inference capacity, letting buyers commit to future token blocks at posted prices and resell unused contracts. The platform offers volume contracts on named models like GLM 5.2 and Kimi K3 Fast, plus floating Top-1 series that settle on the market's leading coding model. Buyers can secure up to 35% off list prices by committing early, with delivery through a single OpenAI-compatible API key.
Funding
Funding not disclosed
Founders
Product
Problem
Reserved AI inference capacity is negotiated privately, with prices varying by buyer and commitments that cannot be resold. Companies that commit forward for discounts get locked into contracts with no exit if usage drops, while GPU providers finance buildouts years ahead without visibility into forward demand or locked-in revenue.
Solution
Touchmark creates an open market for reserved AI inference capacity by putting standardized contracts on a live order book. Buyers purchase fixed token blocks on named open models for specific delivery months at posted prices, earning discounts of up to 35% for early commitment. Providers can sell forward against visible demand, and unused contracts can be relisted for resale on the book. The platform also offers floating Top-1 series that commit to whichever model leads a published coding index, with synthetic units that settle into token allocations at launch.
Target Audience
Primary customers are AI application companies that want to lock in inference capacity at discounted forward prices, and GPU providers or model operators seeking to sell future capacity against visible demand.
Features
- Live two-sided order book with best-ask pricing, bid/ask spreads, and forward curves across delivery months
- Standardized volume contracts for fixed token allocations on models like GLM 5.2 and Kimi K3 Fast, with input and cache-read conversion rates
- Floating Top-1 series that lock in the market's best coding model before it is named, using synthetic units sized on a fixed benchmark set
- Request-for-quote mechanism where buyers post custom terms and providers bid against each other, with quotes visible to the buyer in real time
- Delivery through a single OpenAI-compatible API key, with usage drawing down the contract at pinned per-token rates
- Resale functionality allowing untouched contracts to be relisted on the book before the first token is consumed
- Spot-normalized index tracking forward prices across models and delivery months to benchmark AI capacity as a commodity