InferenceX
中文
MiniMax M3 428B · Chip comparison

MiniMax M3 428B — H200 vs MI325X

Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI325X (AMD CDNA 3) on MiniMax M3 428B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.

AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →

Near the low end of the 27–31 tok/s/user interactivity band, at 27 tok/s/user on MiniMax M3 428B: H200 runs 3948 tok/s/chip at $0.09/M tokens, MI325X runs 999 at $0.31/M. H200 is 256% cheaper per token; H200 delivers 295% more tok/s/chip.

Setting 29 tok/s/user as the target on MiniMax M3 428B, H200 produces 3495 tok/s/chip ($0.10 per million tokens) and MI325X produces 953 ($0.32). H200 is 231% cheaper per token; H200 delivers 267% more tok/s/chip.

At 30 tok/s/user interactivity on MiniMax M3 428B, H200 delivers 2793 tok/s/chip at $0.12 per million tokens; MI325X delivers 919 tok/s/chip at $0.33. H200 is 174% cheaper per token; H200 delivers 204% more tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)

View performance-per-dollar view →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/chip)
H200:3948.0MI325X:999.2
H200:3494.9MI325X:952.8
H200:2793.0MI325X:919.4
Cost ($/M tok)
H200:$0.086MI325X:$0.306
H200:$0.097MI325X:$0.321
H200:$0.121MI325X:$0.332
tok/s/MW
H200:2881739MI325X:591245
H200:2551006MI325X:563803
H200:2038706MI325X:544053
Concurrency
H200:~10MI325X:~3
H200:~8MI325X:~2
H200:~7MI325X:~2

Inference Performance

Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.

Total Tokens per $1 USD (Owning - Neocloud Giant) vs. Interactivity

MiniMax M3 428B • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™

TCO $/chip/hr: VR200: 3.61H100: 1.55H200: 1.59B200: 2.07B300: 2.52GB200: 2.26GB300: 2.79MI300X: 1.16MI325X: 1.32MI355X: 2.09RTX6000PRO: 0.75Jalapeño (Teacup): 1.56

Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model

Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.

Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.

Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.

Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.

Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.

No data available

Please change the model, sequence, precision, date range or chip selection.

Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip