gpt-oss 120B · Chip comparison
gpt-oss 120B — B200 vs H100
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H100 (NVIDIA Hopper) on gpt-oss 120B. 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.
Throughput at 90 tok/s/user on gpt-oss 120B: B200 hits 36,090 tok/s/chip, H100 hits 6,362. Costs per million tokens land at $0.013 and $0.051 respectively. B200 is 74% cheaper per token; B200 delivers 5.7× the tok/s/chip.
B200 / H100 on gpt-oss 120B at 143 tok/s/user: 24,876 / 4,148 tok/s/chip, $0.019 / $0.078 per million tokens. B200 is 75% cheaper per token; B200 delivers 6.0× the tok/s/chip.
Toward the upper edge of the 37–250 tok/s/user interactivity band, at 197 tok/s/user on gpt-oss 120B: B200 runs 17,465 tok/s/chip at $0.028 per million tokens, H100 runs 2,636 at $0.12. B200 is 78% cheaper per token; B200 delivers 6.6× the tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B20036,090H1006,362 | B20024,876H1004,148 | B20017,465H1002,636 |
| Cost ($/M tok) | B200$0.013H100$0.051 | B200$0.019H100$0.078 | B200$0.028H100$0.12 |
| tok/s/MW | B20021,105,097H1004,643,748 | B20014,547,213H1003,027,501 | B20010,213,266H1001,924,056 |
| Concurrency | B200~42H100~32 | B200~32H100~13 | B200~17H100~6 |