Qwen2.5-VL 72B Instruct
qwen/qwen2.5-vl-72b-instruct
Qwen (Alibaba)· 33K ctx
open weights quant: unknown
spread (output, p90 ÷ p10)
— too few providers
$0.800 – $6.88 · 3 providers · max÷min 8.6×
3 providers priced · 4 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 |
|---|---|---|---|---|
NanoGPT router / aggregator | $0.700 | $0.700 | $0.350 | $0.477 |
NovitaAI inference host | $0.800 | $0.800 | — | $0.800no cache price published |
OpenRouter router / aggregator | $0.80 | $1.00 | $0.40 | $0.56 |
Kilo Gateway router / aggregator | $0.80 | $1.00 | $0.40 | $0.56 |
OVHcloud AI Endpoints inference host | $1.01 | $1.01 | — | $1.01no cache price published |
Cortecs router / aggregator | $1.01 | $1.01 | — | $1.01no cache price published |
Alibaba (China)CNfirst-party first-party lab | $2.29 | $6.88 | — | $2.71no cache price published |
Data: models.dev (MIT), snapshot 2026-09-20. Prices are vendor-published; verify before purchase. Methodology.