DeepSeek V3.1 Terminus
deepseek-ai/deepseek-v3.1-terminus
DeepSeek· 128K ctx· mit
open weightstool callstructured quant: unknown
spread (output, p90 ÷ p10)
— too few providers
$1.00 – $1.00 · 3 providers · max÷min 1.0×
3 providers priced · 5 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 | $0.250 | $0.700 | $0.125 | $0.211 |
Alibaba (China)CN first-party lab | $0.27 | $1.00 | — | $0.34no cache price published |
Helicone router / aggregator | $0.27 | $1.00 | $0.22 | $0.30 |
Kilo Gateway router / aggregator | $0.27 | $1.00 | $0.14 | $0.25 |
NovitaAI inference host | $0.27 | $1.00 | $0.14 | $0.25 |
OpenRouter router / aggregator | $0.27 | $1.00 | $0.14 | $0.25 |
SiliconFlow inference host | $0.27 | $1.00 | — | $0.34no cache price published |
Abacus router / aggregator | $0.27 | $1.00 | — | $0.34no cache price published |
Vercel AI Gateway router / aggregator | $0.27 | $1.00 | $0.14 | $0.25 |
Data: models.dev (MIT), snapshot 2026-09-20. Prices are vendor-published; verify before purchase. Methodology.