Model and Modality PlaybooksSeptember 22, 2026Flatkey Team

DeepSeek V4 Flash Pricing (October 2026): Peak, Off-Peak, and Free Access

DeepSeek Flash (now V4.1-Flash-1017) costs $0.30 input and $1.20 output per 1M tokens at peak, and half that off-peak. See the full price table, peak hours, a worked cost example, and what is actually free.

DeepSeek V4 Flash Pricing (October 2026): Peak, Off-Peak, and Free Access

DeepSeek's Flash model costs $0.30 per 1M input tokens and $1.20 per 1M output tokens at peak hours, and exactly half that off-peak ($0.15 input, $0.60 output), with cache-hit input as low as $0.003 per 1M tokens, as of October 25, 2026. The model name you call is now deepseek-flash, which is served by DeepSeek-V4.1-Flash-1017. The old deepseek-v4-flash name still works, but DeepSeek has retired that model and routes those requests to V4.1 Flash at the same Flash price.

There is no permanent free API tier. The DeepSeek chat app is free to use, and API usage is billed against your account balance, with any granted (promotional) balance spent first. This guide covers the full price table, the peak-hour schedule, what a real workload costs, how Flash compares with V4 Pro, and the honest options for trying it for free or close to free. Every number comes from DeepSeek's official pricing page, checked on October 25, 2026.

DeepSeek V4 Flash pricing table

Prices are in USD per 1M tokens on the official DeepSeek API.

	DeepSeek Flash (`deepseek-flash`)	DeepSeek V4 Pro (`deepseek-v4-pro`)
Model version	DeepSeek-V4.1-Flash-1017	DeepSeek-V4-Pro-0813
Input, cache hit (off-peak / peak)	$0.003 / $0.006	$0.022 / $0.044
Input, cache miss (off-peak / peak)	$0.15 / $0.30	$0.66 / $1.32
Output (off-peak / peak)	$0.60 / $1.20	$1.98 / $3.96
Context length	1M tokens	1M tokens
Max output	384K tokens	384K tokens
Vision input	Yes	No
Concurrency limit	2,500	500

Both models support JSON output, tool calls, the Responses API, an Anthropic-compatible endpoint, and chat prefix completion (beta). Fill-in-the-middle (FIM) completion works in non-thinking mode only. DeepSeek also sells a Search/Fetch tool at $0.75 per 1,000 successful requests, which is free inside the Harness, Claude Code, and Codex integrations until billing starts at 00:00 Beijing time on November 1, 2026.

Peak and off-peak hours

Off-peak rates are half of peak rates. According to DeepSeek, 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 all of Saturday and Sunday and the whole of every Chinese public holiday.

Time zone	Peak window 1	Peak window 2
UTC	01:00–04:00	06:00–10:00
Beijing (UTC+8)	09:00–12:00	14:00–18:00
New York (EDT, UTC−4)	21:00–00:00 (previous day)	02:00–06:00
London (GMT, UTC+0 from Oct 25)	01:00–04:00	06:00–10:00

In practice, peak pricing covers the Chinese working day. If your users are in the Americas or Europe, a large share of your daytime traffic already lands in off-peak hours. Batch jobs, evals, and backfills are the easiest to move: schedule them for weekends or outside the two UTC windows and they cost half as much.

What changed: V4 Flash is now V4.1 Flash

Many pages still talk about "DeepSeek V4 Flash" as if it were the current model. As of October 2026 it is not. DeepSeek's API quick start says to use deepseek-flash as the model name, which is now served by DeepSeek-V4.1-Flash-1017. The legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted, but their models are retired, so you get V4.1 Flash output and the Flash price either way.

What this means for you:

Your code keeps working. No request will fail just because it says deepseek-v4-flash.
Your evals may drift. You are now testing a different model checkpoint. Re-run your quality checks if you approved Flash months ago. Our DeepSeek V4 migration checklist covers what to re-test.
Update your config. Switch to deepseek-flash so logs and dashboards show what is actually running. You can track future changes on the DeepSeek API change log.

What a real workload costs

Prices per million tokens are hard to picture, so here is a worked example. Say a support assistant handles 10,000 requests a day. Each request sends a 3,000-token prompt, of which a 2,000-token system prompt and knowledge block is reused (a cache hit), and returns 500 tokens.

That is 20M cache-hit input tokens, 10M cache-miss input tokens, and 5M output tokens per day.

Daily cost	Flash, peak	Flash, off-peak	V4 Pro, peak	V4 Pro, off-peak
Cache-hit input (20M)	$0.12	$0.06	$0.88	$0.44
Cache-miss input (10M)	$3.00	$1.50	$13.20	$6.60
Output (5M)	$6.00	$3.00	$19.80	$9.90
Total per day	$9.12	$4.56	$33.88	$16.94
Total per 30 days	$273.60	$136.80	$1,016.40	$508.20

Three things move this bill the most. Output tokens are the biggest line, so cap max_tokens and ask for short answers where you can. The cache-hit price is about 2% of the cache-miss price, so keep the reusable part of the prompt at the front and identical across requests. And time of day halves everything, so move flexible work off-peak.

Here is the same math as a small Python function you can adapt:

# DeepSeek official prices per 1M tokens, checked October 25, 2026
PRICES = {
    "flash": {"hit": (0.003, 0.006), "miss": (0.15, 0.30), "out": (0.60, 1.20)},
    "pro":   {"hit": (0.022, 0.044), "miss": (0.66, 1.32), "out": (1.98, 3.96)},
}

def daily_cost(model, hit_m, miss_m, out_m, peak=True):
    i = 1 if peak else 0
    p = PRICES[model]
    return hit_m * p["hit"][i] + miss_m * p["miss"][i] + out_m * p["out"][i]

print(round(daily_cost("flash", 20, 10, 5, peak=True), 2))   # 9.12
print(round(daily_cost("flash", 20, 10, 5, peak=False), 2))  # 4.56

How DeepSeek's cache pricing works

DeepSeek caching is automatic: you don't add any flag. Per the DeepSeek context caching guide, each request writes cache units to disk, and a later request gets a cache hit only when its prefix fully matches a stored unit. The response tells you how it went in usage.prompt_cache_hit_tokens and usage.prompt_cache_miss_tokens.

To get more hits:

Put the system prompt, tool definitions, and reference documents first, and keep them byte-for-byte identical.
Put the user's question and anything that changes (timestamps, IDs) last.
For multi-turn chat, append new turns rather than rewriting earlier ones.

If you want a broader view of how caching works across providers, see our prompt caching cost and ROI guide.

Flash vs V4 Pro: when to pay more

Flash costs roughly a quarter of V4 Pro on input and about 30% of it on output. Use Flash by default for chat, extraction, classification, summarization, RAG answers, and agent sub-steps. It also has vision input and a much higher concurrency limit (2,500 vs 500). Move a task to V4 Pro only when your own evals show Flash failing on it, typically on hard multi-step reasoning or large code changes. Our DeepSeek V4 Pro vs Flash comparison and the DeepSeek V4 Pro API pricing guide go deeper.

Is DeepSeek V4 Flash free?

Not on the API. Here is what is and is not free as of October 25, 2026:

Option	Cost	What you get
DeepSeek chat app (web and mobile)	Free	Chat with DeepSeek's current models; not an API
DeepSeek API, granted balance	Free while it lasts	Promotional credit, if your account has any; used before your paid balance
DeepSeek API, paid balance	Pay as you go	The prices in the table above; no subscription
Open-weight checkpoints	Free to download	Self-host on your own GPUs; check the DeepSeek Hugging Face organization for which models and licenses are published

DeepSeek's pricing page states that fees are deducted from your topped-up balance or granted balance, using the granted balance first when you have both. It does not promise a standing free quota for new accounts, so treat any "free DeepSeek API key" offer as a third-party promotion and read its terms. For a wider list of genuinely free options, see our guide to free LLM APIs.

The cheapest paid path is usually good enough for testing: at off-peak prices, a thousand 1,000-token test prompts with 300-token answers cost about $0.33 on Flash.

Call DeepSeek Flash from code

DeepSeek's API is OpenAI-compatible at https://api.deepseek.com, and it also has an Anthropic-format endpoint at https://api.deepseek.com/anthropic (see the Anthropic API guide).

curl https://api.deepseek.com/chat/completions \
  -H "Authorization: Bearer $DEEPSEEK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-flash",
    "messages": [
      {"role": "system", "content": "You are a concise support assistant."},
      {"role": "user", "content": "Summarize our refund policy in two sentences."}
    ],
    "max_tokens": 300
  }'

from openai import OpenAI

client = OpenAI(api_key="YOUR_DEEPSEEK_KEY", base_url="https://api.deepseek.com")
r = client.chat.completions.create(
    model="deepseek-flash",
    messages=[{"role": "user", "content": "Give me three test cases for a date parser."}],
    max_tokens=300,
)
print(r.choices[0].message.content)
print(r.usage)  # check prompt_cache_hit_tokens / prompt_cache_miss_tokens

Using DeepSeek Flash through Flatkey

If you also use Claude, Gemini, Qwen, or GPT models, you can reach DeepSeek through Flatkey with one API key and one balance. As of October 25, 2026, the Flatkey DeepSeek V4.1 Flash model page lists prices from $0.09 per 1M input tokens, $0.0018 per 1M cached tokens, and $0.36 per 1M output tokens, and the legacy deepseek-v4-flash route lists from $0.132 input and $0.396 output. "From" prices are the lowest listed tier, so check the model page for the tier that applies to your traffic.

from openai import OpenAI

client = OpenAI(api_key="YOUR_FLATKEY_KEY", base_url="https://router.flatkey.ai/v1")
r = client.chat.completions.create(
    model="deepseek-v4-1-flash",
    messages=[{"role": "user", "content": "Classify this ticket: 'card charged twice'"}],
)
print(r.choices[0].message.content)

The router quickstart shows how to set this up in a few minutes, and our DeepSeek API pricing overview compares DeepSeek with other low-cost models.

FAQ

How much does DeepSeek V4 Flash cost?

As of October 25, 2026, DeepSeek Flash costs $0.30 per 1M cache-miss input tokens, $0.006 per 1M cache-hit input tokens, and $1.20 per 1M output tokens at peak, and half of that off-peak: $0.15, $0.003, and $0.60.

What are DeepSeek's peak hours?

Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday to Friday, excluding Chinese public holidays. That is 09:00–12:00 and 14:00–18:00 Beijing time. Everything else, including weekends, is off-peak at half price.

Is the DeepSeek API free?

No. The chat app is free, but the API is pay as you go. If your account has a granted balance, DeepSeek uses it before your paid balance.

Is deepseek-v4-flash still available?

The name is still accepted, but the model is retired. Requests to deepseek-v4-flash are served by DeepSeek-V4.1-Flash-1017 and billed at the Flash price. Use deepseek-flash in new code.

What is the context window of DeepSeek Flash?

1M tokens, with up to 384K output tokens, per DeepSeek's pricing page.

Is DeepSeek Flash cheaper than V4 Pro?

Yes. Flash input is $0.30 vs $1.32 per 1M tokens at peak, and output is $1.20 vs $3.96. Flash also has a higher concurrency limit and supports image input.