Alibaba's Qwen3.7-Plus and DeepSeek V4.1-Flash are budget workhorses with 1M-token context. Both have pricing quirks — Qwen raises its rate past 256K tokens, DeepSeek doubles at peak hours — but DeepSeek stays cheaper through both.
DeepSeek V4.1-Flash is cheaper in every case we can model. Off-peak it is $0.15/$0.60 per 1M; at its $0.30/$1.20 peak it is still below Qwen3.7-Plus's base $0.40/$1.60, and far below Qwen's $1.20/$4.80 rate for 256K–1M-token prompts.
| Qwen3.7-Plus | DeepSeek V4.1-Flash | |
|---|---|---|
| Provider | Alibaba | DeepSeek |
| Tier | Budget | Budget |
| Input / 1M tokens | $0.40 | $0.15 |
| Output / 1M tokens | $1.60 | $0.60 |
| Cache hit / 1M | $0.04 | $0.003 |
| Cache write / 1M | — | — |
| Context window | 1M | 1M |
| Notes | Balanced Qwen workhorse; base tier list price shown (0-256K tokens); 256K-1M tokens: $1.20/$4.80; both bands currently carry a limited-time 20% discount | Released Sep 10, 2026, replacing V4-Flash & V4-Flash-Vision-Exp at a lower price; off-peak rate shown. Peak (01:00-04:00 & 06:00-10:00 UTC weekdays) is 2x: $0.30/$1.20 |
DeepSeek V4.1-Flash is 2.7× cheaper on input tokens ($0.15 vs $0.40 per 1M). DeepSeek V4.1-Flash is 2.7× cheaper on output tokens ($0.60 vs $1.60 per 1M).
| Workload | Qwen3.7-Plus | DeepSeek V4.1-Flash | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $14.40 | $5.40 | DeepSeek V4.1-Flash saves $9.00 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $13.28 | $4.98 | DeepSeek V4.1-Flash saves $8.30 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $0.54 | $0.16 | DeepSeek V4.1-Flash saves $0.38 |
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 are on Alibaba Cloud already, want one rate regardless of time of day, or your evals favour Qwen's output for your language or domain.
cost is the deciding factor — it is cheaper at peak, off-peak, short and long prompts, and its cache hits cost under a tenth of Qwen's.
The Qwen3.7-Plus rate in the table covers prompts up to 256K tokens. From 256K to 1M it rises to $1.20 input and $4.80 output per 1M, triple the base. DeepSeek V4.1-Flash keeps one rate across its full 1M window, so long-document workloads widen the gap considerably.
DeepSeek doubles prices on weekdays 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays. Unlike some rivals, that doubling still leaves V4.1-Flash below Qwen3.7-Plus's base rate, so peak hours affect your margin but not which model is cheaper.
Cache hits are $0.003 per 1M off-peak on DeepSeek versus $0.04 on Qwen. For chat apps and agents that resend history, that difference compounds quickly.
Yes, on input, output and cache hits, at both off-peak and peak rates.
Yes. Prompts between 256K and 1M tokens are billed at $1.20 input and $4.80 output per 1M.
Yes, both list 1M tokens.
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.