Kimi K2.5/K2.6/K2.7-Code 1T — B300 vs GB300 NVL72
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB300 NVL72 (NVIDIA Blackwell) 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.
B300 / GB300 NVL72 on Kimi K2.5/K2.6/K2.7-Code 1T at 61 tok/s/user: 2968 / 10636 tok/s/chip, $0.21 / $0.06 per million tokens. GB300 NVL72 is 251% cheaper per token; GB300 NVL72 delivers 258% more tok/s/chip.
Around the middle of the 24–173 tok/s/user interactivity band, at 99 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: B300 runs 1731 tok/s/chip at $0.36/M tokens, GB300 NVL72 runs 1785 at $0.36/M. Cost per token is essentially tied; GB300 NVL72 delivers 3% more tok/s/chip.
Setting 136 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, B300 produces 783 tok/s/chip ($0.80 per million tokens) and GB300 NVL72 produces 876 ($0.73). GB300 NVL72 is 9% cheaper per token; GB300 NVL72 delivers 12% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:2968.0GB300 NVL72:10636.5 | B300:1731.4GB300 NVL72:1785.3 | B300:783.1GB300 NVL72:876.2 |
| Cost ($/M tok) | B300:$0.212GB300 NVL72:$0.060 | B300:$0.363GB300 NVL72:$0.359 | B300:$0.802GB300 NVL72:$0.732 |
| tok/s/MW | B300:1562093GB300 NVL72:5017202 | B300:911253GB300 NVL72:842121 | B300:412168GB300 NVL72:413316 |
| Concurrency | B300:~22GB300 NVL72:~1330 | B300:~8GB300 NVL72:~74 | B300:~3GB300 NVL72:~28 |
Total Tokens per $1 USD (Owning - Neocloud Giant) vs. Interactivity
Kimi K2.5/2.6/2.7-Code 1T • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™
Kimi K2.5/2.6/2.7-Code 1T • 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
Interactivity is the rate at which a single user receives generated tokens while the model streams its answer — how quickly new words appear on screen. Higher values feel snappier; operators trade it against batch throughput.
How many total tokens (input + output) one US dollar of infrastructure spend buys, priced with the all-in hourly ownership cost of a Neocloud Giant operator. It is the reciprocal of cost per token, so higher means cheaper.
Formula: tok/$ = (total tok/s/chip × 3,600) ÷ all-in cost per chip-hour ($)