Poko API

Clear model pricing for agentic video work.

Every paid plan can spend Poko credits on GPT through Codex and on Poko Engine (Pulse and Flow). Claude models need BYOK, which starts on Plus for One-Time Access and on Pro for subscriptions.Poko Engine is Poko's managed model layer for Pulse and Flow. GPT on Poko credits is billed at the published OpenAI catalog rate for the selected model.

Poko Engine rates

USD per 1M tokens

Cached input refers to cache reads.

ModelInputOutputCached inputBest for

Poko Pulse

Quality default

$0.90$2.90$0.30For full video builds and complex, multi-step work.

Poko Pulse Priority

Low-latency routing

$1.00$4.00$0.40Pulse routed to lower-latency providers for time-sensitive work.

Poko Flow

Fast iteration

$1.30$5.30$0.26For fast edits, iteration, and short-turnaround tasks.

Poko Flow Priority

Fast + low latency

$2.00$8.00$0.40Flow routed to lower-latency providers for urgent iteration.

One balance

Poko credits work with GPT and Poko Engine.

Every paid plan can spend Poko credits on GPT through Codex and on Poko Engine (Pulse and Flow). Claude models need BYOK, which starts on Plus for One-Time Access and on Pro for subscriptions.The selected model's published rate determines the charge.

Bring your own key

Your provider, your bill.

Requests made with a supported personal provider key do not spend Poko credits. BYOK starts on Plus for One-Time Access and on Pro for subscriptions. Claude models on new purchases need BYOK.

Built for video agents

Creative work, metered clearly.

Poko Motion uses its workflow layer to turn source material into editable video projects while keeping model usage visible and controllable.

Getting started

Choose Pulse for quality. Choose Flow for speed.

Start with Pulse for full builds. Use Flow when you want fast, economical iteration. Priority variants keep the same model family while prioritizing processing.