Kimi K3 Fast
- Token weights
- Output1.00×Input0.20×Cached0.02×
- Service levels
- Availability≥ 99.0%TTFT≤ 1,000 msp95Speed≥ 250 t/sp05
- Duration
- 1 month
- Provider
- Wafer
For illustrative purposes only
4 trillion plus tokens of RFQ volume
A new instrument for buying
and selling reserved inference.
For illustrative purposes only
Specify the performance you need. Providers commit to meeting it.
Multiple providers bid on the same order. Trades typically clear 10-60% below retail. Buyers never pay transaction fees.
Your capacity isn’t locked in. Use it or sell it to another buyer.
01Reserve the baseline. Open a request for quotes to reserve capacity. Customize your volume, terms and timeline.
02Reserve the peaks. Purchase throughput up to 3 hours ahead of demand, with the same SLA-backed protections.
Our vision
AI is a commodity.Its financial infrastructurehasn’t caught up.
Capacity is just our starting point.