consumer_gpu
NVIDIA RTX 5090
Facts
- VRAM
- 32 GB 0 Source
- Reference rental (CNY/h)
- 2.4 ¥/h 0 Source
- Rental source
- 润云 2.4 / 共绩 2.5 / 并行智算云 2.98 (2026)link ↗
- Note
- 包月约 1482 元(润云)
cost = p50 ms ÷ 1000 × rental ÷ 3600 × 1000
Sources · 0 sources
This record has no sources yet; every field is treated as unverified.
Latency ranking on this card
| Model | p50 ms | Cost per 1k inferences ¥ | samples |
|---|---|---|---|
| Isaac GR00T N1.7 NVIDIA | — | — no latency data | No data |
| GR00T N1.6 3B NVIDIA | — | — no latency data | No data |
| LingBot-VLA 2.0 蚂蚁灵波科技 (Ant Group) | — | — no latency data | No data |
| GO-1 (Genie Operator-1) 智元机器人 (AgiBot) | — | — no latency data | No data |
| π0 (pi-zero) Physical Intelligence | — | — no latency data | No data |
| π0.5 Physical Intelligence | — | — no latency data | No data |
| SmolVLA Hugging Face (LeRobot) | — | — no latency data | No data |
| OpenVLA Stanford 等 | — | — no latency data | No data |
| X-VLA 学术团队 (ICLR 2026) | — | — no latency data | No data |
| Octo UC Berkeley 等 | — | — no latency data | No data |
| 星海图具身基础模型(WRC 开源,名称待定) 星海图 (Galaxea) | — | — no latency data | No data |
No measurements on this card yet. The crowd CLI (sinan-bench) will feed this table; until then rows show “—”.