DeepSeek V4 Pro vs Qwen3.8 Max
DeepSeek V4 Pro (DeepSeek) costs $1.32 per million input tokens and $3.96 per million output tokens; Qwen3.8 Max (Alibaba Cloud) costs $2.00 and $6.00. DeepSeek V4 Pro is 1.5× cheaper for a typical 3:1 input-to-output mix. DeepSeek V4 Pro has the larger context window (1M vs 992K tokens).
| DeepSeek V4 Pro | Qwen3.8 Max | |
|---|---|---|
| Provider | DeepSeek | Alibaba Cloud |
| Input / 1M tokens | $1.32 | $2.00 |
| Output / 1M tokens | $3.96 | $6.00 |
| Cached input / 1M | $0.044 | $0.25 |
| Blended (3:1) | $1.98 | $3.00 |
| Context window | 1M | 992K |
| Max output | 393K | 131K |
| Vision | No | Yes |
| Tool use | Yes | Yes |
| Reasoning | Yes | Yes |
| Listed since | Jun 20, 2026 | Aug 16, 2026 |
Cost per 1,000 requests
| Workload | DeepSeek V4 Pro | Qwen3.8 Max |
|---|---|---|
| Chatbot reply2,000 input + 500 output tokens | $4.62 | $7.00 |
| RAG / search answer8,000 input + 700 output tokens | $13.33 | $20.20 |
| Long-document summary100,000 input + 2,000 output tokens | $139.92 | $212.00 |
Blended price history
Blended = (3 × input + output) ÷ 4, per million tokens. Each line starts when the model first appeared in the price list.
The companies
List prices for each provider's own API (USD), taken daily from LiteLLM's public model price list. Batch, caching and committed-use discounts aren't reflected. Price alone says nothing about output quality; check benchmarks and your own evaluations before switching.