Llama 3.3 70B · Chip comparison
Llama 3.3 70B — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) on Llama 3.3 70B. 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.
B200 / H200 on Llama 3.3 70B at 45 tok/s/user: 5,994 / 2,978 tok/s/chip, $0.080 / $0.11 per million tokens. B200 is 30% cheaper per token; B200 delivers 101% more tok/s/chip.
Around the middle of the 14–138 tok/s/user interactivity band, at 76 tok/s/user on Llama 3.3 70B: B200 runs 3,981 tok/s/chip at $0.12 per million tokens, H200 runs 1,930 at $0.18. B200 is 31% cheaper per token; B200 delivers 106% more tok/s/chip.
Setting 108 tok/s/user as the target on Llama 3.3 70B, B200 produces 2,352 tok/s/chip ($0.20 per million tokens) and H200 produces 1,152 ($0.29). B200 is 31% cheaper per token; B200 delivers 104% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B2005,994H2002,978 | B2003,981H2001,930 | B2002,352H2001,152 |
| Cost ($/M tok) | B200$0.080H200$0.11 | B200$0.12H200$0.18 | B200$0.20H200$0.29 |
| tok/s/MW | B2003,505,000H2002,173,427 | B2002,328,088H2001,408,963 | B2001,375,171H200840,884 |
| Concurrency | B200~32H200~32 | B200~28H200~11 | B200~9H200~8 |