DeepSeek V4.1 Flash API Pricing: Peak vs Off-Peak Cost (2026)
Updated October 17, 2026 · Prices checked: October 17, 2026
This page shows what DeepSeek V4.1 Flash costs on DeepSeek's own API: the off-peak and peak price per 1M tokens, when peak hours fall in US time zones, what the October 17 "1017" update changed, and how the cost compares with similar models. Every price links to an official page. Anything we could not confirm is marked Not verified.
Quick answer
- Off-peak: $0.15 per 1M input tokens (cache miss), $0.003 per 1M input tokens (cache hit), $0.60 per 1M output tokens.
- Peak: double the off-peak rate: $0.30 input (cache miss), $0.006 input (cache hit), $1.20 output, per 1M tokens.
- Peak hours: 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday. In Pacific time until October 31, 2026 that is 6 pm–9 pm and 11 pm–3 am, Sunday to Thursday nights. Weekends and Chinese public holidays are off-peak all day (DeepSeek pricing page).
DeepSeek V4.1 Flash API price per 1M tokens
API model name: deepseek-flash. Current version: DeepSeek-V4.1-Flash-1017 (DeepSeek pricing page).
| Token type | Off-peak (per 1M) | Peak (per 1M) | Source |
|---|---|---|---|
| Input (cache miss) | $0.15 | $0.30 | DeepSeek pricing page |
| Input (cache hit) | $0.003 | $0.006 | DeepSeek pricing page |
| Output | $0.60 | $1.20 | DeepSeek pricing page |
DeepSeek bills tokens × price and states that it may adjust prices. These rates took effect at 04:00 UTC on September 10, 2026 (DeepSeek V4.1 Flash announcement).
Peak vs off-peak hours (in your time zone)
DeepSeek defines peak hours in UTC: 01:00–04:00 and 06:00–10:00 UTC, Monday through Friday, excluding Chinese public holidays. All other hours are off-peak, including weekends and Chinese public holidays in full (DeepSeek pricing page). DeepSeek does not list the holiday dates on that page, so check with DeepSeek for specific days. The US times below are our conversion of the official UTC window.
US clocks change on Sunday, November 1, 2026. From that day the US times are one hour earlier. The UTC and Beijing times do not change.
| Window | UTC (Mon–Fri) | Pacific (PDT) | Mountain (MDT) | Central (CDT) | Eastern (EDT) | Beijing (Mon–Fri) |
|---|---|---|---|---|---|---|
| Peak 1 | 01:00–04:00 | 6 pm–9 pm, Sun–Thu | 7 pm–10 pm, Sun–Thu | 8 pm–11 pm, Sun–Thu | 9 pm–12 am, Sun–Thu | 09:00–12:00 |
| Peak 2 | 06:00–10:00 | 11 pm–3 am, Sun–Thu nights | 12 am–4 am, Mon–Fri | 1 am–5 am, Mon–Fri | 2 am–6 am, Mon–Fri | 14:00–18:00 |
| Window | UTC (Mon–Fri) | Pacific (PST) | Mountain (MST) | Central (CST) | Eastern (EST) | Beijing (Mon–Fri) |
|---|---|---|---|---|---|---|
| Peak 1 | 01:00–04:00 | 5 pm–8 pm, Sun–Thu | 6 pm–9 pm, Sun–Thu | 7 pm–10 pm, Sun–Thu | 8 pm–11 pm, Sun–Thu | 09:00–12:00 |
| Peak 2 | 06:00–10:00 | 10 pm–2 am, Sun–Thu nights | 11 pm–3 am, Sun–Thu nights | 12 am–4 am, Mon–Fri | 1 am–5 am, Mon–Fri | 14:00–18:00 |
Peak time adds up to 7 hours a day × 5 days = 35 hours out of 168 hours a week, about 20.8% of the week. Arizona does not use daylight time, so it follows the Mountain standard (MST) row all year.
What changed on October 17 (Flash-1017)
- On October 17, 2026 DeepSeek upgraded DeepSeek V4.1 Flash to DeepSeek-V4.1-Flash-1017. DeepSeek says the API calling method and pricing remain unchanged, and
deepseek-flashnow serves the new version (DeepSeek change log). - The legacy names
deepseek-v4-flashanddeepseek-v4-flash-vision-expare still accepted. Those models are retired, so requests are served by DeepSeek-V4.1-Flash-1017 and billed at the Flash price (DeepSeek pricing page). - The 1017 weights are published under the MIT License (model card). DeepSeek says the update also shortens task runtime and output length (change log).
Specs that affect cost
- Context length: 1M tokens. Maximum output: 384K tokens (DeepSeek pricing page).
- Cache hits are much cheaper: $0.003 vs $0.15 per 1M input tokens off-peak, so repeated prompt prefixes (agents, long system prompts) cost far less (DeepSeek pricing page).
- Features: JSON output, tool calls, Responses API, Anthropic API format and vision (image input). FIM completion works in non-thinking mode only (DeepSeek pricing page).
- Concurrency limit: 2500 (DeepSeek pricing page).
- Model size: 552B-parameter mixture of experts, 8B active parameters for input and 16B for output (DeepSeek news).
- Search and Fetch APIs are billed separately at $0.75 per 1,000 successful requests each. Calls made through DeepSeek Harness, Claude Code or Codex are free until per-request billing starts at 00:00 Beijing time on November 1, 2026 (DeepSeek pricing page).
Worked cost examples
Fixed examples using the official rates above. "Cache miss" input is billed at the full input rate; "cache hit" input at the cache-hit rate.
| Usage | All off-peak | All peak |
|---|---|---|
| Small: 10M input (all cache miss) + 2M output | 10 × $0.15 + 2 × $0.60 = $1.50 + $1.20 = $2.70 | 10 × $0.30 + 2 × $1.20 = $3.00 + $2.40 = $5.40 |
| Agent: 100M input (80M cache hit, 20M cache miss) + 10M output | 80 × $0.003 + 20 × $0.15 + 10 × $0.60 = $0.24 + $3.00 + $6.00 = $9.24 | 80 × $0.006 + 20 × $0.30 + 10 × $1.20 = $0.48 + $6.00 + $12.00 = $18.48 |
| Large: 500M input (all cache miss) + 50M output | 500 × $0.15 + 50 × $0.60 = $75.00 + $30.00 = $105.00 | 500 × $0.30 + 50 × $1.20 = $150.00 + $60.00 = $210.00 |
If your traffic is spread evenly over the week, about 20.8% of it lands in peak hours. Real bills depend on when you send requests and how much of your input hits the cache.
Cost vs other models
| Model | Input / 1M | Cached input / 1M | Output / 1M | Pricing type | Context | Example: 10M in + 2M out | Source |
|---|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash, off-peak | $0.15 | $0.003 | $0.60 | Peak / off-peak | 1M | 10 × $0.15 + 2 × $0.60 = $2.70 | DeepSeek pricing page |
| DeepSeek V4.1 Flash, peak | $0.30 | $0.006 | $1.20 | Peak / off-peak | 1M | 10 × $0.30 + 2 × $1.20 = $5.40 | DeepSeek pricing page |
| DeepSeek V4-Pro, off-peak | $0.66 | $0.022 | $1.98 | Peak / off-peak | Not verified | 10 × $0.66 + 2 × $1.98 = $10.56 | DeepSeek pricing page |
| DeepSeek V4-Pro, peak | $1.32 | $0.044 | $3.96 | Peak / off-peak | Not verified | 10 × $1.32 + 2 × $3.96 = $21.12 | DeepSeek pricing page |
| GLM-5.3-Flash (Z.ai) | $0.15 | $0.03 | $0.50 | One rate (cached input storage listed as limited-time free) | 1M (model page) | 10 × $0.15 + 2 × $0.50 = $2.50 | Z.ai pricing page |
| Ultramarine 1 Flash (Paper Instruments) | $0.15 | $0.03 | $0.50 | One rate | Not verified | 10 × $0.15 + 2 × $0.50 = $2.50 | Ultramarine page |
| Muse Spark Open (Together AI) | $1.25 | $0.15 | $4.25 | One rate | 1,048,576 (official docs) | 10 × $1.25 + 2 × $4.25 = $21.00 | Together AI announcement. More: Muse Spark Open self-host vs hosted cost |
The DeepSeek pricing page shows the 1M context length in a cell shared by both models, so we do not state V4-Pro's context separately. DeepSeek's September 10 announcement said deepseek-v4-pro requests would route to V4.1 Flash from September 14, but the change log now says V4 Pro stays available with billing unchanged, and the pricing page still lists it. Paper Instruments says Ultramarine 1 Flash API pricing is "the same as GLM-5.3 Flash"; its page says the model weights are coming soon.
Who should choose which
Choose DeepSeek V4.1 Flash off-peak if…
- You can schedule batch or agent jobs outside the peak windows above, including weekends.
- Your prompts reuse long prefixes, so much of your input is billed at the $0.003 cache-hit rate.
Choose GLM-5.3-Flash or Ultramarine 1 Flash if…
- You want one price at any hour. Both list $0.15 input and $0.50 output per 1M tokens, with no peak window on their pricing pages.
Choose Muse Spark Open if…
- You need to self-host so prompts stay on your own machines, or you want that specific model. See our Muse Spark Open self-host vs hosted cost page for hardware needs and hosted prices.
Other hosts
Third-party gateways and inference hosts also serve DeepSeek V4.1 Flash at their own rates. Those are not DeepSeek's prices, change often, and are not included in the tables above.
FAQ
How much does DeepSeek V4.1 Flash cost per 1M tokens?
As of October 17, 2026, DeepSeek lists off-peak prices of $0.15 per 1M input tokens (cache miss), $0.003 per 1M input tokens (cache hit) and $0.60 per 1M output tokens. Peak prices are double: $0.30, $0.006 and $1.20.
When are DeepSeek off-peak hours?
Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday through Friday, excluding Chinese public holidays. Every other hour is off-peak, including weekends and Chinese public holidays. In US Pacific time that is 6 pm–9 pm and 11 pm–3 am, Sunday to Thursday nights, until October 31, 2026. From November 1, 2026 it is one hour earlier: 5 pm–8 pm and 10 pm–2 am.
What is DeepSeek V4.1 Flash 1017?
DeepSeek-V4.1-Flash-1017 is the October 17, 2026 update of DeepSeek V4.1 Flash. The API model name is still deepseek-flash, and DeepSeek says the calling method and pricing are unchanged.
What is the context window of DeepSeek V4.1 Flash?
DeepSeek lists a 1M-token context length and a maximum output of 384K tokens.
Is DeepSeek V4.1 Flash cheaper than GLM-5.3-Flash?
It depends on when you send requests and how often your input hits the cache. For 10M input tokens with no cache hits plus 2M output tokens, DeepSeek V4.1 Flash costs $2.70 off-peak or $5.40 at peak, and GLM-5.3-Flash costs $2.50 at its single listed rate.
Can I still use the deepseek-v4-flash model name?
Yes. DeepSeek still accepts deepseek-v4-flash, but the old model is retired. Requests are served by DeepSeek-V4.1-Flash-1017 and billed at the Flash price.