Phase 8 · SLOs & Cost

The Cost of a Token

GPU-hour → tokens → margin, from first principles.

0/4 · 530 XP

You’re previewing the curriculum. Sign in (top right) to track progress, take knowledge checks, and unlock the verifiers.

Tasks

Read “LLM Inference Economics from First Principles”
The best cost-per-token derivation published. Also skim Epoch AI's inference-economics piece.
read+80 XPresource ↗
Study InferenceMAX's methodology
Pareto frontiers and $/M-token across hardware — the reference for how cost claims get made honestly.
read+50 XPresource ↗
Build the cost model for YOUR fleet
Amortized hardware + power → $/M tokens at measured throughput, per model and quantization. This spreadsheet is interview gold for a founder.
build+150 XP
VERIFIED: publish “How we cut cost per token at Portal Labs”
The founder story with real numbers. Your single best piece of evidence.
write+250 XPauto-verified