DeepSeek-V4.1-Flash 是深度求索 2026.9.10 发布的原生多模态 MoE 大模型，百万 token 上下文，全新非对称架构，推理吞吐更高、KV 缓存成本大幅降低，综合 Agent、推理性能优于 V4-Pro，API 支持 OpenAI/Anthropic 双协议。
| Providers | Credit report | Credit | Region | Formats | Price | TTFT | Throughput | 7-day success rate | 7-day calls | Availability (24h) | Status | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| - | Guangzhou | Input: ¥0.74/ 1M tokensOutput: ¥2.96/ 1M tokens | 931 ms | 105 tokens/s | 99.98% | 320.6M | In service | |||||
| View report | 4.5 | Guangzhou | Input: ¥0.8/ 1M tokensOutput: ¥3.2/ 1M tokens | 2114 ms | 18 tokens/s | 99.33% | 15.5K | In service | ||||
| View report | 4.5 | Guangzhou | Input: ¥1.2/ 1M tokensOutput: ¥4.8/ 1M tokens | 1027 ms | 136 tokens/s | 99.79% | 240.1K | In service | ||||
| View report | 4.5 | Guangzhou | Input: ¥1.3/ 1M tokensOutput: ¥5.2/ 1M tokens | 1158 ms | 134 tokens/s | 99.64% | 273.9K | In service | ||||
| View report | 4.5 | Guangzhou | Input: ¥1.5/ 1M tokensOutput: ¥6/ 1M tokens | 1739 ms | 130 tokens/s | 99.91% | 270.5K | In service |
No more
