DeepSeek V4 Flash 0731 API Pricing
DeepSeek · context window 1.3M · prices updated 2026-09-09
Where DeepSeek V4 Flash 0731 sits on price
At $0.18 per million output tokens, DeepSeek V4 Flash 0731 is a budget tier model — 12.2× cheaper than the median output price of $2.20, which puts it in the cheapest 10% of everything we track. It sits 2th cheapest of the 17 DeepSeek models we track. Output is only 2.8× the input price, an unusually flat spread: long prompts hurt about as much as long answers, so prompt size is where the savings are.
Cached input is 75% cheaper. On an input-heavy workload you need roughly a 33% cache-hit rate to take 25% off the input line — reachable for chatbots and agents that resend the same system prompt and history.
What real workloads cost on DeepSeek V4 Flash 0731
These three workloads are the ones teams actually run on a budget-priced model — a $0.18/1M model is not bought for the same job as one ten times the price.
| Workload | Per request | Per month |
|---|---|---|
| High-volume classification — 200K requests/day (400 in / 20 out) Short output, huge volume: the input price dominates the bill. | $0 | $178 |
| Small app — 1K requests/day (500 in / 150 out) The baseline most side projects actually run at. | $0.0001 | $1.78 |
| Long-context analysis — 200 runs/day (250K in / 4K out) Only possible on large-context models; input cost is nearly the whole bill. | $0.017 | $102 |
Model your exact traffic in the API cost calculator — it preloads DeepSeek V4 Flash 0731 with caching and batch options.
DeepSeek V4 Flash 0731 price history
DeepSeek last cut DeepSeek V4 Flash 0731's output price on Sep 8, 2026 (-36% per output token), from $0.14/$0.28 to $0.065/$0.18 per 1M in/out. That is one of 11 repricings in the 38 days we have tracked it — worth knowing before you hard-code today's rate into a budget.
| Date | Input /1M | Output /1M | Output change |
|---|---|---|---|
| Aug 2, 2026 (tracking began) | $0.09 | $0.18 | — |
| Aug 10, 2026 | $0.08 | $0.18 | 0% |
| Aug 14, 2026 | $0.14 | $0.28 | +56% |
| Aug 21, 2026 | $0.08 | $0.18 | -36% |
| Aug 24, 2026 | $0.14 | $0.28 | +56% |
| Aug 26, 2026 | $0.06 | $0.12 | -57% |
| Aug 28, 2026 | $0.05 | $0.1 | -17% |
| Aug 29, 2026 | $0.045 | $0.09 | -10% |
| Aug 30, 2026 | $0.065 | $0.18 | +100% |
| Sep 6, 2026 | $0.05 | $0.1 | -44% |
| Sep 7, 2026 | $0.14 | $0.28 | +180% |
| Sep 8, 2026 | $0.065 | $0.18 | -36% |
Change-points from our daily price snapshots (tracking since Aug 2, 2026; intraday moves between snapshots are not captured).
DeepSeek V4 Flash 0731 vs Llama Guard 4 12B
The closest-priced alternative from another vendor is Llama Guard 4 12B (Meta) — priced within a rounding error on output, with $0.115 more per million input tokens. When two models land this close on price, the decision is quality on your own workload, not the price sheet: run 50 real requests through both and compare.
See Llama Guard 4 12B pricing →Cheaper alternatives
More DeepSeek models
Frequently asked questions
› How much does the DeepSeek V4 Flash 0731 API cost?
DeepSeek V4 Flash 0731 costs $0.065 per million input tokens and $0.18 per million output tokens, with cached input at $0.016 per million (75% cheaper). That works out to roughly 5,555,556 output tokens per dollar.
› What does the high-volume classification workload cost on DeepSeek V4 Flash 0731?
High-volume classification — 200K requests/day (400 in / 20 out) costs about $0 per request and $178 per month on DeepSeek V4 Flash 0731. Short output, huge volume: the input price dominates the bill.
› What does it cost to fill DeepSeek V4 Flash 0731's 1.3M context window?
Sending 1.3M of input in a single request costs $0.0852 at $0.065 per million tokens — before any output. With prompt caching that same fill drops to about $0.021 on repeat requests. This is why large context windows are cheap to advertise and expensive to actually use.
› Is DeepSeek V4 Flash 0731 worth the price?
DeepSeek V4 Flash 0731 is in the cheapest end of the market — cheaper than 95% of the models we track. At this price the question is not cost but capability: it is a good fit for classification, extraction, routing and bulk summarisation, and a poor fit for multi-step reasoning where a wrong answer costs more than the tokens saved.