All model and GPU pairings
MiniMax M3 428BNVIDIA Blackwell

Running MiniMax M3 on B200

Quick answer

MiniMax M3 sustains 7,193 tokens/s per GPU on B200 at 50 tokens/s per user, which works out to $0.067 per million tokens at hyperscaler pricing, served by vLLM. Fastest measured TTFT: 0.0 ms; fastest TPOT: 0.0 ms (each the best across all configs, not one run).

Benchmarked configs

268

Serving engines

Dynamo vLLM, TRTLLM, vLLM

Precisions

fp4, fp8

Run dates

2026-06-232026-08-20

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/s8,377$0.057vLLMfp8
50 tok/s7,193$0.067vLLMfp8
75 tok/s8,656$0.056vLLMfp4
100 tok/s6,511$0.074vLLMfp4
150 tok/s4,736$0.10vLLMfp4
200 tok/s3,699$0.13vLLMfp4

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$1.73$0.067
Neocloud$2.07$0.080
Retail$2.60$0.10

Frequently asked questions

How fast is MiniMax M3 on B200?
At an interactivity target of 50 tokens/s per user on a single-turn chat workload (8k input / 1k output), B200 sustains 7,193 tokens/s per GPU serving MiniMax M3 with vLLM in FP8. Peak measured throughput across all configs is 48,610 tokens/s per GPU.
How much does it cost to serve MiniMax M3 on B200?
$0.067 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 B200?
The runs behind this page used Dynamo vLLM, TRTLLM, vLLM in FP4, FP8, including disaggregated prefill and multi-node serving. 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 B200 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 2026-08-20. The same derivation powers the InferenceX overview leaderboard.

Explore the data