DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows....
| Field | Per token (USD) | Per million |
|---|---|---|
| prompt | 0.000000048 | $0.048 |
| completion | 0.000000144 | $0.144 |
| input_cache_read | 0.000000012 | $0.012 |