Kimi K2 Instruct
moonshotai/kimi-k2-instruct
Moonshot AI· 131K ctx· other
open weightstool callstructured quant: unknown
spread (output, p90 ÷ p10)
— too few providers
$2.29 – $3.00 · 3 providers · max÷min 1.3×
3 providers priced · 3 offerings
Cheapest / median / spread use providers that serve the model on their own capacity: first-party lab, hyperscaler, inference host.
Blended $/M assumptions
blended = 90.9% input × [(1 − 70% cache) × input + 70% cache × cache-read] + 9.1% × output + cache writes (30% of uncached input) where a write price is published
A provider that publishes no cache price is blended with cache share 0 (all input at the input price) and labelled “no cache price published” — never penalised against identical in/out prices.
| Provider | Input | Output | Cache rd | Blended |
|---|---|---|---|---|
NanoGPTfp4 router / aggregator | $0.40 | $1.80 | $0.20 | $0.40 |
Alibaba (China)CN first-party lab | $0.57 | $2.29 | — | $0.73no cache price published |
NovitaAI inference host | $0.57 | $2.30 | — | $0.73no cache price published |
Jiekou.AI reseller | $0.57 | $2.30 | — | $0.73no cache price published |
Hugging Face inference host | $1.00 | $3.00 | — | $1.18no cache price published |
Data: models.dev (MIT), snapshot 2026-09-20. Prices are vendor-published; verify before purchase. Methodology.