# DeepSeek: DeepSeek V3.2 — price and measured latency by host

Updated July 25, 2026. Served by 14 providers; 6 of them measured here.

Latency and uptime are measured by llmlatency.dev probes. Prices are published by the providers and read via OpenRouter on 2026-07-25 — not measured.

| Host | Input $/M | Output $/M | Context | Fastest p50 | From | Uptime |
| --- | --- | --- | --- | --- | --- | --- |
| siliconflow | $0.259 | $0.420 | 163,840 | 82 ms | us-central | 100% |
| deepinfra | $0.260 | $0.380 | 163,840 | 226 ms | us-central | 99.9% |
| novita | $0.269 | $0.400 | 163,840 | 71 ms | us-central | 100% |
| friendli | $0.500 | $1.500 | 163,840 | 185 ms | us-central | 100% |
| google | $0.560 | $1.680 | 163,840 | 40 ms | us-central | 100% |
| sambanova | $3.000 | $4.500 | 32,768 | 24 ms | ap-tokyo | 99.8% |

Source: https://llmlatency.dev/model/deepseek-deepseek-v3.2 · CC BY 4.0