Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) 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 100–300 tok/s/user interactivity band, at 150 tok/s/user on MiniMax M3 428B: GB300 NVL72 runs 34992 tok/s/chip at $0.02/M tokens, MI355X runs 28972 at $0.01/M. MI355X is 28% cheaper per token; GB300 NVL72 delivers 21% more tok/s/chip.
Setting 200 tok/s/user as the target on MiniMax M3 428B, GB300 NVL72 produces 24623 tok/s/chip ($0.03 per million tokens) and MI355X produces 19317 ($0.02). MI355X is 21% cheaper per token; GB300 NVL72 delivers 27% more tok/s/chip.
At 250 tok/s/user interactivity on MiniMax M3 428B, GB300 NVL72 delivers 17236 tok/s/chip at $0.04 per million tokens; MI355X delivers 8573 tok/s/chip at $0.05. GB300 NVL72 is 31% cheaper per token; GB300 NVL72 delivers 101% more tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:34992.4MI355X:28972.1 | GB300 NVL72:24623.5MI355X:19317.1 | GB300 NVL72:17236.4MI355X:8573.3 |
| Cost ($/M tok) | GB300 NVL72:$0.018MI355X:$0.014 | GB300 NVL72:$0.026MI355X:$0.022 | GB300 NVL72:$0.037MI355X:$0.049 |
| tok/s/MW | GB300 NVL72:16505842MI355X:13862238 | GB300 NVL72:11614857MI355X:9242634 | GB300 NVL72:8130384MI355X:4102080 |
| Concurrency | GB300 NVL72:~34MI355X:~18 | GB300 NVL72:~20MI355X:~12 | GB300 NVL72:~23MI355X:~4 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr:
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.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
1 The ATOM engine is promising, however it has yet to serve production tokens. It is still in its infant stage.
MiniMax M3 428B 8K / 1K Total Tokens per $1 TCO vs. Interactivity