Ssylphx.ai
Pricing

One flat price per tier

Pick a tier and get one flat rate — and it stays flat no matter what the engine does to earn your answer. Councils, invisible background quality checks, routing between lanes: none of them is a separate charge. Public rates arrive with our launch.

Four tiers. Same contract, different ceiling.

Every tier runs the same engine and the same lanes, with full capability — vision, tools, streaming, English and 中文. What a tier changes is the quality ceiling you're paying for, never the API surface.

Lumen Flash

Cheapest lane that clears the bar. Maximum thrift.

sylphx/lumen-flash
PriceComing with public launch
Built forInteractive apps, chat surfaces, high-volume light work
How it picks a laneTakes the cheapest lane that clears the quality bar, accepting a wide band below the predicted best.
Default

Lumen

The default. Best value in the family.

sylphx/lumen
PriceComing with public launch
Built forEveryday agentic work
How it picks a laneTakes the cheapest lane near the predicted best, with a tight tolerance.

Lumen Pro

The predicted best. You pay for the ceiling.

sylphx/lumen-pro
PriceComing with public launch
Built forComplex agentic work, planning, hard reasoning
How it picks a laneTakes the predicted-best lane; cost only breaks near-ties.

Lumen Ultra

Council on demand

Zero tolerance — plus a council when it's hard.

sylphx/lumen-ultra
PriceComing with public launch
Built forResearch, critique, the decisions that matter
How it picks a laneTakes the predicted best with zero tolerance. When a request looks hard or unfamiliar, it convenes a multi-lane council and an independent judge picks the served answer.

Three flat rates. However many steps it takes to earn your answer.

What the flat rate actually covers

Three numbers, fixed per tier — one for input tokens, one for output tokens, one for web search. No stacked per-mechanism fees, no surprise line items for extra compute.

Flat input rate

One fixed price per million input tokens, set per tier. It never changes with how a request is routed or how many lanes it touches on the way to an answer.

Flat output rate

One fixed price per million output tokens, set per tier — the same whether the answer came from a single lane or a full council deliberation.

Flat web-search rate

One fixed price per web search, set per tier. Only the searches your request actually triggers are billed, at that one rate, with nothing stacked on top.

Internal steps are never a line item

Routing between lanes, invisible background quality checks, and Ultra's council deliberation are how a tier earns its answer — not add-ons you pay for. However many internal steps a request takes, the three flat rates above are the whole bill.

Public rates land with our launch: one flat input rate, one flat output rate, and one flat web-search rate, each fixed per tier. Every tier includes full capability regardless of price. These prices aren't hidden or placeholder — they're genuinely still being set, and we'll publish every rate in full the day pricing goes live.