DeepSeek V4 Pro pricing
DeepSeek V4 Pro, Flagship-tier from DeepSeek, costs $1.32 per 1M input tokens and $3.96 per 1M output tokens.
Price your own workload against DeepSeek V4 Pro in the calculator →
Specs and pricing
| Provider | DeepSeek |
| Tier | Flagship |
| Context window | 1M tokens |
| Max output | 384K tokens |
| Input price | $1.32 / 1M tokens |
| Output price | $3.96 / 1M tokens |
| Cache read | $0.044 / 1M tokens |
| Modalities | text |
| Released | 2026-04 |
Note. Peak cache-miss rate, which is this dataset's standard tier. DeepSeek's own page states off-peak rates are half the peak rates, with peak hours 01:00-04:00 and 06:00-10:00 UTC Monday to Friday: off-peak is $0.66 input / $1.98 output, cache-hit input $0.022. The off-peak half-price is a time-of-day discount and is excluded here for the same reason batch and cached-input discounts are.
What it costs at realistic volumes
Assuming a 4:1 input:output token ratio, a common shape for chat and RAG-style workloads:
| Volume (4:1 in:out ratio) | Input cost | Output cost | Total / month |
|---|---|---|---|
| 1,000,000 in / 250,000 out | $1.32 | $0.99 | $2.31 |
| 10,000,000 in / 2,500,000 out | $13.20 | $9.90 | $23.10 |
| 100,000,000 in / 25,000,000 out | $132.00 | $99.00 | $231.00 |
Cheaper alternatives
Same tier (Flagship), a different provider, and a lower blended cost at the 4:1 ratio used above:
| Model | Provider | $/1M in | $/1M out | At the 4:1 ratio |
|---|---|---|---|---|
| Qwen3 235B A22B Instruct 2507 | Novita AI | $0.09 | $0.58 | 89.8% cheaper blended |
| GPT-OSS 120B | Groq | $0.15 | $0.60 | 87% cheaper blended |
| Qwen3 235B-A22B (2507) | Parasail | $0.14 | $0.80 | 85.3% cheaper blended |
| Llama 4 Maverick | Together AI | $0.27 | $0.85 | 79.1% cheaper blended |
| Llama 4 Maverick Instruct | Novita AI | $0.27 | $0.85 | 79.1% cheaper blended |