FlexAI · The clearest published attempt to make inference pricing checkable rather than trusted — they write down the rule that sets each rate and ship calculators that show where per-token billing stops being the cheap option.

One OpenAI-compatible key that starts cheap and scales into hardware you reserve. Token Factory serves 20+ open-weight models — text, vision, image, video, speech — billed per token, per image, per generated video or per minute. When per-token billing stops making sense you move the same workload onto dedicated endpoints priced per GPU-hour and metered per minute, or onto AI Factory, which is their platform deployed on your own hardware (VPC, on-prem or air-gapped). Managed fine-tuning, distributed training and an Agent SDK (in trial) sit on the same account. The pitch underneath is hardware indifference: the same job runs on NVIDIA, AMD or TPU without a refactor.

Sector

Inference

Stage

Seed

On the roster since

15 Sept 2026