Kimi K3 2.8T · Chip comparison
Kimi K3 2.8T — GB300 NVL72 vs H200
Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) 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 →
Near the low end of the 2–423 tok/s/user interactivity band, at 107 tok/s/user on Kimi K3 2.8T: GB300 NVL72 runs 11,814 tok/s/chip at $0.054 per million tokens, H200 runs 70 at $4.85. GB300 NVL72 is 99% cheaper per token; GB300 NVL72 delivers 169× the tok/s/chip.
Setting 212 tok/s/user as the target on Kimi K3 2.8T, GB300 NVL72 produces 7,811 tok/s/chip ($0.082 per million tokens) and H200 produces 70 ($4.85). GB300 NVL72 is 98% cheaper per token; GB300 NVL72 delivers 112× the tok/s/chip.
At 318 tok/s/user interactivity on Kimi K3 2.8T, GB300 NVL72 delivers 2,982 tok/s/chip at $0.22 per million tokens; H200 delivers 70 tok/s/chip at $4.85. GB300 NVL72 is 96% cheaper per token; GB300 NVL72 delivers 43× the tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL7211,814H20070 | GB300 NVL727,811H20070 | GB300 NVL722,982H20070 |
| Cost ($/M tok) | GB300 NVL72$0.054H200$4.85 | GB300 NVL72$0.082H200$4.85 | GB300 NVL72$0.22H200$4.85 |
| tok/s/MW | GB300 NVL725,572,559H20051,001 | GB300 NVL723,684,297H20051,001 | GB300 NVL721,406,440H20051,001 |
| Concurrency | GB300 NVL72~48H200~1 | GB300 NVL72~25H200~1 | GB300 NVL72~14H200~1 |