Kimi K3 repriced across 4 hosts: Sail Research input down 61%, Fireworks (US region) up 36% on all rates
Published 30 September 2026
Price move, per 1M tokens
- Sail Research · input↓ 61%$0.809 → $0.318
- Sail Research · output↓ 38%$12.75 → $7.94
- Sail Research · cache readunchanged$0.40 → $0.40
- InferenceNet · input↓ 53%$0.40 → $0.19
- InferenceNet · output↑ 22%$9.00 → $11.00
- InferenceNet · cache read↓ 55%$0.40 → $0.18
- Fireworks (US region) · input↑ 36%$3.30 → $4.50
- Fireworks (US region) · output↑ 36%$16.50 → $22.50
- Fireworks (US region) · cache read↑ 36%$0.33 → $0.45
- Morph · input↓ 5%$1.23 → $1.17
- Morph · output↑ 27%$10.70 → $13.55
- Morph · cache readunchanged$0.29 → $0.29
| host | rate | was | now | change |
|---|---|---|---|---|
| Sail Research | input | $0.809 | $0.318 | ↓ 61% |
| Sail Research | output | $12.75 | $7.94 | ↓ 38% |
| Sail Research | cache read | $0.40 | $0.40 | — |
| InferenceNet | input | $0.40 | $0.19 | ↓ 53% |
| InferenceNet | output | $9.00 | $11.00 | ↑ 22% |
| InferenceNet | cache read | $0.40 | $0.18 | ↓ 55% |
| Fireworks (US region) | input | $3.30 | $4.50 | ↑ 36% |
| Fireworks (US region) | output | $16.50 | $22.50 | ↑ 36% |
| Fireworks (US region) | cache read | $0.33 | $0.45 | ↑ 36% |
| Morph | input | $1.23 | $1.17 | ↓ 5% |
| Morph | output | $10.70 | $13.55 | ↑ 27% |
| Morph | cache read | $0.29 | $0.29 | — |
Sail Research input: ↓ 61%InferenceNet input: ↓ 53%InferenceNet output: ↑ 22%Fireworks (US region) all rates: ↑ 36%Morph input: ↓ 5%Morph output: ↑ 27%
If you buy
Fireworks buyers now pay more on every rate for this model, so re-check your budget and whether a lighter model covers the same work before your next top-up.
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 on openrouter.ai · we published this on 30 September 2026.
read the original at openrouter.ai ↗