gpt-oss 120B — Jalapeño vs MI325X
Head-to-head AI inference benchmark comparison of Jalapeño (OpenAI Jalapeño) and MI325X (AMD CDNA 3) 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 100 tok/s/user on gpt-oss 120B: Jalapeño hits 59656 tok/s/chip, MI325X hits 956. Per-million costs land at $0.01 and $0.32 respectively. Jalapeño is 5305% cheaper per token; Jalapeño delivers 6140% more tok/s/chip.
Jalapeño / MI325X on gpt-oss 120B at 102 tok/s/user: 59545 / 839 tok/s/chip, $0.01 / $0.36 per million tokens. Jalapeño is 6046% cheaper per token; Jalapeño delivers 6996% more tok/s/chip.
Toward the upper edge of the 98–107 tok/s/user interactivity band, at 105 tok/s/user on gpt-oss 120B: Jalapeño runs 59379 tok/s/chip at $0.01/M tokens, MI325X runs 618 at $0.49/M. Jalapeño is 8219% cheaper per token; Jalapeño delivers 9504% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | Jalapeño:59656.2MI325X:956.1 | Jalapeño:59545.3MI325X:839.1 | Jalapeño:59378.5MI325X:618.3 |
| Cost ($/M tok) | Jalapeño:$0.006MI325X:$0.320 | Jalapeño:$0.006MI325X:$0.364 | Jalapeño:$0.006MI325X:$0.494 |
| tok/s/MW | Jalapeño:53027756MI325X:565717 | Jalapeño:52929129MI325X:496508 | Jalapeño:52780913MI325X:365830 |
| Concurrency | Jalapeño:~1993MI325X:~5 | Jalapeño:~1954MI325X:~4 | Jalapeño:~1896MI325X:~4 |