DeepSeek's mid-tier V4-Pro and Alibaba's flagship Qwen3.8-Max both offer 1M-token windows at low prices. Their pricing works differently: DeepSeek varies by hour, Alibaba keeps a single tier across the window.
Off-peak, DeepSeek V4-Pro is about 3× cheaper than Qwen3.8-Max on input and output. In DeepSeek's weekday peak window its rates double, cutting the lead to about 1.5×. Qwen3.8-Max is a flagship-tier model with one price at any hour and any prompt length up to 1M tokens.
| DeepSeek V4-Pro | Qwen3.8-Max | |
|---|---|---|
| Provider | DeepSeek | Alibaba |
| Tier | Mid-tier | Flagship |
| Input / 1M tokens | $0.66 | $2.00 |
| Output / 1M tokens | $1.98 | $6.00 |
| Cache hit / 1M | $0.022 | $0.17 |
| Cache write / 1M | — | — |
| Context window | 1M | 1M |
| Notes | A planned Sep 14, 2026 reroute to V4.1-Flash rates was reversed after user pushback; billing unchanged since the Aug 13, 2026 GA build. Off-peak rate shown; peak (01:00-04:00 & 06:00-10:00 UTC weekdays) is 2x: $1.32/$3.96 | New Alibaba flagship (Aug 3, 2026); 2.4T MoE |
DeepSeek V4-Pro is 3.0× cheaper on input tokens ($0.66 vs $2.00 per 1M). DeepSeek V4-Pro is 3.0× cheaper on output tokens ($1.98 vs $6.00 per 1M).
| Workload | DeepSeek V4-Pro | Qwen3.8-Max | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $21.12 | $64.00 | DeepSeek V4-Pro saves $42.88 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $21.38 | $64.80 | DeepSeek V4-Pro saves $43.42 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $0.64 | $2.31 | DeepSeek V4-Pro saves $1.67 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
you can run most traffic off-peak, your work is text-only, and a mid-tier model passes your evals.
you want a flagship-tier model with the same price at any hour, or a 1M-token free quota to evaluate on Alibaba Cloud's international region.
DeepSeek doubles V4-Pro's rates from 01:00–04:00 and 06:00–10:00 UTC on weekdays. If a large share of your traffic lands in that window — European mornings, for example — calculate a blended rate. Even at peak, V4-Pro stays cheaper than Qwen3.8-Max's $2.00/$6.00 on standard rates, and the same holds for cached input: V4-Pro's peak cache hit is double its off-peak $0.022, still well under Qwen's $0.17.
Alibaba lists qwen3.8-max as a single price tier from 0 to 1M input tokens, billing all tokens in a request at that rate, with the same price for thinking and non-thinking modes. That predictability is worth something for customer-facing products where you can't control when users arrive.
DeepSeek lists V4-Pro as text-only, thinking-by-default, with up to 384K output tokens and a 500-request concurrency limit. Check Alibaba's console for Qwen3.8-Max limits in your region before sizing a launch.
Yes: about 3× cheaper off-peak and about 1.5× at peak on input and output.
No time-of-day pricing is listed. Alibaba prices qwen3.8-max as a single tier up to 1M input tokens.
Alibaba lists a 1M-token free quota for qwen3.8-max in its Singapore region.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.