Unlimited tokens.
Powerful open models.

Pay monthly for GPU time, not tokens

Running the best of open source
HOW UNLIMITED WORKS

Unlimited tokens. Without the all-day price.

Reserve a daily eight-hour window and pay for the hours you need.

Your daily window

Choose one of three eight-hour blocks. Your window opens at the same time every day.

Unlimited while you’re in

Use every model included in your plan. No token bundles to ration or per-token bills.

Only pay for your hours

Choose one block for eight hours daily. Your monthly subscription renews with the same hours.

It’s genuinely unlimited

No per-token charges during your daily eight-hour window.

Choose one daily eight-hour block. Core is $23 per month for any block.

Flagship

Included Models

  • Kimi K3

Unlimited tokens

$199 / month

Monthly billing only

Your hours: Europe · 8 AM–4 PM daily

Frontier

Included Models

  • GLM 5.3
  • MiniMax M3

Unlimited tokens

$71 / month

Monthly billing only

Your hours: Europe · 8 AM–4 PM daily

Core

Included Models

  • DeepSeek V4.1 Flash
  • DeepSeek V4.1 Flash Uncensored

Unlimited tokens

$23 / month

Plus applicable tax · renews monthly

Your hours: Europe · 8 AM–4 PM daily

Built for agentic workflows and coding harnesses, with stable prompt prefixes for better cache use. Requests run only during your selected hours. Fair use allows one active request per account, so subagents must send requests sequentially. Token speed varies with available hardware, current electricity costs, context, and caching. Automatic speed throttling may apply when fewer than 80% of input tokens are served from cache. Abusive use may be restricted. Access may also be slowed or paused when processing capacity is limited; if a plan reaches its processing capacity, access may pause until the next billing period. Publicly sharing API keys, reselling access, and unlawful use are prohibited. Access covers the models included in your chosen plan. We replace included models only with newer, more efficient options of equal or better capability. Logos do not imply affiliation. Hermes has been tested. Other clients may need an OpenAI-compatible adapter; native Anthropic Messages is supported. Subscriptions renew monthly, plus applicable tax. Cancel future renewals from your dashboard. Polar is our merchant of record; cancellation and refund terms apply. All use must comply with our Terms of Use.

Keep the tools you already code with

Connect Hermes or an OpenAI-compatible client.

Illustration of a developer laptop connected by cable to inference servers