{"unit":"USD per 1M tokens (request + response, chars/4 rounded down)","models":[{"model":"mamba","reserved_usd_per_mtok":2.0,"spot_usd_per_mtok":0.8},{"model":"nemotron","reserved_usd_per_mtok":5.0,"spot_usd_per_mtok":2.0}],"spot":{"how":"send {\"tier\": \"spot\"} in the request body","semantics":"preemptible — first-party training preempts spot jobs; a preempted call returns 503 (not billed). No SLA on latency."},"privacy":"private by ownership: prompts run on hardware we own, no third-party API. NOT TEE-attested — for hardware attestation use the Phala TEE lane.","job_fair":"the 'GPU Inference / Reasoning' skill (5 USDC) clears through this lane"}