Kimi K3 API pricing
Kimi K3 by Sference costs $3.00 per 1M input tokens and $15.00 per 1M output tokens (1M in + 1M out = $18.00).
- API model id
Kimi K3- Input / 1M
- $3.00
- Output / 1M
- $15.00
- Cached input / 1M
- $0.45
- Context window
- —
- Notes
- Sference serverless (European GPUs), USD per 1M tokens at realtime priority; rates 'confirmed per model when your account is provisioned'.
- Source
- https://sference.com/
- Provider type
- Inference host (HQ: United Kingdom)
- Regions
- EU
- EU data residency
- yes — Home page: 'Every request runs on European GPUs' (low-latency realtime inference on European GPUs). Security page: production runs in a dedicated AWS VPC in eu-central-1 (Frankfurt). Sference is a trading name of Neural Compute Ltd, London, England (company no. 17195150). Serverless rates are published in USD at realtime priority and 'confirmed per model when your account is provisioned'.
- Last verified
- 2026-10-19
Cost estimate
| Provider | Model | Cost |
|---|---|---|
| Sference | Kimi K3 | $6.75 |
Uses standard list prices (no caching, batch or off-peak discounts). € and CHF rows are priced in that currency by the provider and not converted, so compare them with care.
Other Sference models
- GLM 5.3: $1.20 / $4.20
- DeepSeek V4.1 Flash (1017): $0.50 / $1.50
- GLM 5.3 Flash: $0.20 / $0.60
- DeepSeek V4 Flash (0731): $0.28 / $0.56