gpt-oss 120B · Chip comparison

gpt-oss 120B — H100 vs MI355X

Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and MI355X (AMD CDNA 4) 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.

H100 / MI355X on gpt-oss 120B at 91 tok/s/user: 6,311 / 26,707 tok/s/chip, $0.051 / $0.016 per million tokens. MI355X is 70% cheaper per token; MI355X delivers 4.2× the tok/s/chip.

Around the middle of the 38–250 tok/s/user interactivity band, at 144 tok/s/user on gpt-oss 120B: H100 runs 4,116 tok/s/chip at $0.079 per million tokens, MI355X runs 16,480 at $0.025. MI355X is 68% cheaper per token; MI355X delivers 4.0× the tok/s/chip.

Setting 197 tok/s/user as the target on gpt-oss 120B, H100 produces 2,636 tok/s/chip ($0.12 per million tokens) and MI355X produces 10,740 ($0.039). MI355X is 69% cheaper per token; MI355X delivers 4.1× 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.)

View performance per dollar →

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)
H1006,311MI355X26,707
H1004,116MI355X16,480
H1002,636MI355X10,740
Cost ($/M tok)
H100$0.051MI355X$0.016
H100$0.079MI355X$0.025
H100$0.12MI355X$0.039
tok/s/MW
H1004,606,456MI355X12,778,496
H1003,004,142MI355X7,885,227
H1001,924,056MI355X5,138,603
Concurrency
H100~32MI355X~34
H100~13MI355X~13
H100~6MI355X~6