Arbitrage · DeepSeek V3.25 providers · $0.140 → $0.750 · 81% spread
Cheapest Provider for DeepSeek V3.2
MoE reasoning model · rewrites cost structure of frontier-tier inference.
Cheapest input
DeepSeek (direct)
$0.140/M
Off-peak pricing 50% off · cache read $0.014
Fastest
H
Hyperbolic
60 tok/s
Middle ground · flexible
Savings calculator
Save 81%
vs Together AI at $0.750/M input. For 100M tokens/mo, that is $61/mo saved by routing to DeepSeek (direct).
Sorted by input price
All 5 providers
| Provider | In $/M | Out $/M |
|---|---|---|
| DeepSeek (direct)Winner | $0.140 | $0.280 |
H Hyperbolic | $0.250 | $1.00 |
 OpenRouter | $0.300 | $1.20 |
 Fireworks AI | $0.560 | $1.68 |
 Together AI | $0.750 | $2.40 |
Notes: DeepSeek (direct) · Off-peak pricing 50% off · cache read $0.014 ; Hyperbolic · Middle ground · flexible ; OpenRouter · Multi-provider routing ; Fireworks AI · 4x markup vs DeepSeek direct ; Together AI · Premium · US data residency
Frequently Asked Questions
DeepSeek (direct) at $0.140/M input and $0.280/M output. That is 81% cheaper than Together AI. Off-peak pricing 50% off · cache read $0.014.