Kimi K3 cut across 2 hosts, by up to 22% at Relace (output), with cache read down 26% there
Published 19 September 2026
Price move, per 1M tokens
- Relace · input↓ 19%$2.10 → $1.70
- Relace · output↓ 22%$10.95 → $8.50
- Relace · cache read↓ 26%$0.23 → $0.17
- Wafer · input↓ 16%$2.50 → $2.10
- Wafer · outputunchanged$10.95 → $10.95
- Wafer · cache read↓ 16%$0.25 → $0.21
| host | rate | was | now | change |
|---|---|---|---|---|
| Relace | input | $2.10 | $1.70 | ↓ 19% |
| Relace | output | $10.95 | $8.50 | ↓ 22% |
| Relace | cache read | $0.23 | $0.17 | ↓ 26% |
| Wafer | input | $2.50 | $2.10 | ↓ 16% |
| Wafer | output | $10.95 | $10.95 | — |
| Wafer | cache read | $0.25 | $0.21 | ↓ 16% |
Relace output: ↓ 22%Wafer input and cache read: ↓ 16%
The model
Kimi K3open the model →
Open weights2779.9B parameters1049K context25 hosts
intelligence12th of 168 writing18th of 168 coding6th of 168 agents10th of 55
Kimi K3 is Moonshot AI's massive-scale mixture-of-experts model with 2.8 trillion total parameters and 104 billion active per word. It excels at reasoning and web-development tasks, and handles up to one million tokens in a single request, though its weights carry commercial restrictions.
Source
Read from the hosts' own published rates · we published this on 19 September 2026.