Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on GLM 5.3 744B. 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 →
Setting 125 tok/s/user as the target on GLM 5.3 744B, B200 produces 10435 tok/s/chip ($0.05 per million tokens) and MI355X produces 10441 ($0.04). MI355X is 15% cheaper per token; throughput per chip is essentially tied.
At 165 tok/s/user interactivity on GLM 5.3 744B, B200 delivers 9714 tok/s/chip at $0.05 per million tokens; MI355X delivers 7041 tok/s/chip at $0.06. B200 is 20% cheaper per token; B200 delivers 38% more tok/s/chip at this point.
B200 posts 6482 tok/s/chip for $0.07 per million tokens at 204 tok/s/user on GLM 5.3 744B; MI355X posts 3359 tok/s/chip for $0.12. B200 is 67% cheaper per token; B200 delivers 93% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:10435.2MI355X:10440.8 | B200:9713.6MI355X:7040.7 | B200:6482.4MI355X:3359.4 |
| Cost ($/M tok) | B200:$0.046MI355X:$0.040 | B200:$0.049MI355X:$0.059 | B200:$0.074MI355X:$0.124 |
| tok/s/MW | B200:6102454MI355X:4995578 | B200:5680489MI355X:3368764 | B200:3790906MI355X:1607386 |
| Concurrency | B200:~26MI355X:~7 | B200:~47MI355X:~4 | B200:~8MI355X:~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.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
GLM5.2/GLM5.3 744B 8K / 1K Total Tokens per $1 TCO vs. Interactivity