GLM 5.3 744B — B200 vs TPU7x Performance per Dollar
Cost per million tokens of B200 (NVIDIA Blackwell) versus TPU7x (Google Ironwood) on GLM 5.3 744B. Large-hyperscaler-volume ownership TCO normalized by total 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 →
TPU7x edges B200 at 98 tok/s/user on GLM 5.3 744B — $0.04 per million tokens versus $0.06, a 49% cost-per-token gap.
Push GLM 5.3 744B to 151 tok/s/user and B200 lands at $0.07 per million tokens against TPU7x's $0.05 — TPU7x pulls ahead by 53%.
B200: $0.11 per million tokens. TPU7x: $0.07. Both at 203 tok/s/user on GLM 5.3 744B, with TPU7x 70% cheaper. (Numbers reflect the default agentic-traces · fp8 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 · TPU7x $1.03/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:$0.056TPU7x:$0.038 | B200:$0.073TPU7x:$0.048 | B200:$0.114TPU7x:$0.067 |
| Concurrency | B200:~14TPU7x:~12 | B200:~8TPU7x:~6 | B200:~5TPU7x:~4 |