Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB300 NVL72 (NVIDIA Blackwell) on Kimi K3 2.8T. 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 →
B300 / GB300 NVL72 on Kimi K3 2.8T at 57 tok/s/user: 5615 / 12022 tok/s/chip, $0.11 / $0.05 per million tokens. GB300 NVL72 is 109% cheaper per token; GB300 NVL72 delivers 114% more tok/s/chip.
Around the middle of the 9–202 tok/s/user interactivity band, at 106 tok/s/user on Kimi K3 2.8T: B300 runs 3879 tok/s/chip at $0.16/M tokens, GB300 NVL72 runs 7717 at $0.08/M. GB300 NVL72 is 95% cheaper per token; GB300 NVL72 delivers 99% more tok/s/chip.
Setting 154 tok/s/user as the target on Kimi K3 2.8T, B300 produces 2069 tok/s/chip ($0.30 per million tokens) and GB300 NVL72 produces 5432 ($0.12). GB300 NVL72 is 157% cheaper per token; GB300 NVL72 delivers 163% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:5615.2GB300 NVL72:12022.2 | B300:3879.1GB300 NVL72:7716.7 | B300:2068.8GB300 NVL72:5431.8 |
| Cost ($/M tok) | B300:$0.112GB300 NVL72:$0.053 | B300:$0.162GB300 NVL72:$0.083 | B300:$0.303GB300 NVL72:$0.118 |
| tok/s/MW | B300:2955360GB300 NVL72:5670831 | B300:2041612GB300 NVL72:3639938 | B300:1088859GB300 NVL72:2562157 |
| Concurrency | B300:~9GB300 NVL72:~48 | B300:~5GB300 NVL72:~33 | B300:~2GB300 NVL72:~25 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
