Kimi K3 2.8T — B200 vs B300 Performance per Dollar
Cost per million tokens of B200 (NVIDIA Blackwell) versus B300 (NVIDIA Blackwell) on Kimi K3 2.8T. Owning-hyperscaler TCO normalized by output tokens — performance per dollar across LLM workloads. Pick the more cost-efficient SKU at every target interactivity level. 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 →
B200: $0.12 per million tokens. B300: $0.11. Both at 57 tok/s/user on Kimi K3 2.8T, with B300 7% cheaper.
Around the middle of the 8–202 tok/s/user interactivity band — at 105 tok/s/user — B200 runs $0.24 per million tokens on Kimi K3 2.8T while B300 runs $0.16. B300 is the cheaper choice by 47%.
On Kimi K3 2.8T at 154 tok/s/user, the per-million math comes out to $0.36 for B200 and $0.30 for B300; B300 delivers 20% more output per dollar. (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.)
Chip pricing (owning hyperscaler): B200 $1.73/chip/hr · B300 $2.26/chip/hr. Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | B200:$0.120B300:$0.112 | B200:$0.236B300:$0.161 | B200:$0.364B300:$0.303 |
| Concurrency | B200:~14B300:~9 | B200:~6B300:~5 | B200:~3B300:~2 |