Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) on Kimi K2.5/K2.6/K2.7-Code 1T. 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.
B200 / H200 on Kimi K2.5/K2.6/K2.7-Code 1T at 45 tok/s/user: 1588 / 821 tok/s/chip, $0.30 / $0.41 per million tokens. B200 is 36% cheaper per token; B200 delivers 93% more tok/s/chip.
Around the middle of the 30–92 tok/s/user interactivity band, at 61 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: B200 runs 1178 tok/s/chip at $0.41/M tokens, H200 runs 673 at $0.50/M. B200 is 23% cheaper per token; B200 delivers 75% more tok/s/chip.
Setting 77 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, B200 produces 824 tok/s/chip ($0.58 per million tokens) and H200 produces 530 ($0.64). B200 is 10% cheaper per token; B200 delivers 56% more tok/s/chip. (Numbers reflect the default 8k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:1587.6H200:821.0 | B200:1177.6H200:672.9 | B200:824.5H200:529.5 |
| Cost ($/M tok) | B200:$0.303H200:$0.413 | B200:$0.408H200:$0.504 | B200:$0.583H200:$0.640 |
| tok/s/MW | B200:928398H200:599258 | B200:688654H200:491168 | B200:482136H200:386505 |
| Concurrency | B200:~32H200:~17 | B200:~17H200:~10 | B200:~10H200:~6 |
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
Kimi K2.5/2.6/2.7-Code 1T 8K / 1K Total Tokens per $1 TCO vs. Interactivity