Muse Spark Open 816B — B200 vs TPU7x
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and TPU7x (Google Ironwood) on Muse Spark Open 816B. 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 →
Setting 179 tok/s/user as the target on Muse Spark Open 816B, B200 produces 20031 tok/s/chip ($0.02 per million tokens) and TPU7x produces 17723 ($0.02). TPU7x is 49% cheaper per token; B200 delivers 13% more tok/s/chip.
At 264 tok/s/user interactivity on Muse Spark Open 816B, B200 delivers 11778 tok/s/chip at $0.04 per million tokens; TPU7x delivers 14152 tok/s/chip at $0.02. TPU7x is 102% cheaper per token; TPU7x delivers 20% more tok/s/chip at this point.
B200 posts 5205 tok/s/chip for $0.09 per million tokens at 348 tok/s/user on Muse Spark Open 816B; TPU7x posts 9123 tok/s/chip for $0.03. TPU7x is 194% cheaper per token; TPU7x delivers 75% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:20030.6TPU7x:17722.9 | B200:11778.0TPU7x:14151.8 | B200:5204.5TPU7x:9122.6 |
| Cost ($/M tok) | B200:$0.024TPU7x:$0.016 | B200:$0.041TPU7x:$0.020 | B200:$0.092TPU7x:$0.031 |
| tok/s/MW | B200:11713829TPU7x:14683445 | B200:6887744TPU7x:11724803 | B200:3043581TPU7x:7558056 |
| Concurrency | B200:~12TPU7x:~11 | B200:~9TPU7x:~9 | B200:~5TPU7x:~5 |