All model and GPU pairings
MiniMax M3 428BNVIDIA Hopper

Running MiniMax M3 on H200

Quick answer

MiniMax M3 sustains 1,719 tokens/s per GPU on H200 at 50 tokens/s per user, which works out to $0.20 per million tokens at hyperscaler pricing, served by vLLM. Fastest measured TTFT: 0.1 ms; fastest TPOT: 0.0 ms (each the best across all configs, not one run).

Benchmarked configs

114

Serving engines

vLLM

Precisions

fp8

Run dates

2026-06-132026-08-12

Throughput at every interactivity target

Serving is a trade-off: push more concurrent users through a GPU and each user's tokens arrive slower. The ladder below reads the measured frontier at each per-user speed target on a single-turn chat workload (8k input / 1k output), using the best engine and precision at that point.

Per-user targetTokens/s per GPU$ / 1M tokensEnginePrecision
30 tok/s1,920$0.18vLLMfp8
50 tok/s1,719$0.20vLLMfp8
75 tok/s1,423$0.24vLLMfp8
100 tok/s1,179$0.29vLLMfp8
150 tok/s897$0.38vLLMfp8
200 tok/s754$0.45vLLMfp8

What serving actually costs

Converting the 50 tokens/s per user operating point to $ per million total tokens across rental pricing tiers from the SemiAnalysis AI Cloud TCO model.

Pricing tier$ / GPU / hr$ / 1M tokens
Hyperscaler$1.22$0.20
Neocloud$1.59$0.26
Retail$2.05$0.33

Frequently asked questions

How fast is MiniMax M3 on H200?
At an interactivity target of 50 tokens/s per user on a single-turn chat workload (8k input / 1k output), H200 sustains 1,719 tokens/s per GPU serving MiniMax M3 with vLLM in FP8. Peak measured throughput across all configs is 4,467 tokens/s per GPU.
How much does it cost to serve MiniMax M3 on H200?
$0.20 per million total tokens at hyperscaler $/GPU/hr pricing, at 50 tokens/s per user. Neocloud and retail rental tiers are tabulated above; slower interactivity targets lower the cost further.
Which serving engines run MiniMax M3 on H200?
The runs behind this page used vLLM in FP8. Engines are rebuilt and re-benchmarked continuously, so the best config can change between visits.
How are these MiniMax M3 numbers measured?
Every number is measured on real H200 hardware by the InferenceX fleet, sweeping concurrency on a single-turn chat workload (8k input / 1k output) to trace the throughput-versus-interactivity frontier; the newest run landed on 2026-08-12. The same derivation powers the InferenceX overview leaderboard.

Explore the data