AgentX · MiniMax-M3.1-Flash FP4 · MI355X
Total tokens per $ of TCO*, against p90 interactivity.
* GPU cost: $1.50 per MI355X GPU-hour, owned at large hyperscaler volume[1].
[1] SemiAnalysis, “AMD Instinct MI355X Specs, Pricing & AI Inference Benchmarks,” InferenceX, inferencex.semianalysis.com/chips/mi355x. Hourly rate from the SemiAnalysis AI Cloud TCO Model; accessed October 2026.