← Compare all APIs

DeepSeek-V4.1-Flash-1017 API prices

Compare routes under the same recorded conditions.

Official pricing source ↗

Channel choices

API: deepseek/deepseek-flash · USD · Standard · Precision unstated · 1M · Region not specified · Peak billing window

Reference: 1M input + 1M output tokens; no cache or extras. Not a bill.

Only 1 channel in this scope. Configuration is incomplete; equivalent replacement and savings are unverified.

View recorded configuration conditions

Context limit: 1000000

Input/output limits: Unrecorded / Unrecorded

Regions: unspecified

Billing scope: Unrecorded

Parameter declarations: reasoning_effort, response_format, thinking, tools

Native tool declarations: Unrecorded

Implicit caching declaration: Unrecorded

Account restriction: Unrecorded

Subroute label: Unrecorded

These are recorded source fields; actual capability, performance and account access remain unverified.

Change billing and configuration scope

DeepSeek

Subroute unrecorded

Direct billing channel

$1.50

1M input + 1M output token subtotal

Input / output per 1M tokens: $0.30 / $1.20

Lowest reference rate in this group; verify output and access.

Use this endpoint
Official API docs ↗

Create a provider API key and set the local environment variable DEEPSEEK_API_KEY, then run the example below.

Request URL
https://api.deepseek.com/chat/completions
Connection settings
Base URL: https://api.deepseek.com
Catalog model: deepseek-flash
Request API model: deepseek-flash
Protocol: openai
API key env: DEEPSEEK_API_KEY
Docs: https://api-docs.deepseek.com/
curl
curl "https://api.deepseek.com/chat/completions" -H "Authorization: Bearer $DEEPSEEK_API_KEY" -H "Content-Type: application/json" -d '{
  "model": "deepseek-flash",
  "messages": [
    {
      "role": "user",
      "content": "Read this support ticket and return a JSON object with \"summary\", \"urgency\" (low, medium, or high), and \"next_step\". Use null for missing details.\n\nTicket: I paid for the Pro plan yesterday, but my account still shows Free. My team needs access before tomorrow’s demo."
    }
  ]
}'
Prompt for your coding agent
Use the OpenAI-compatible API at https://api.deepseek.com with model "deepseek-flash".  Read the API key from the DEEPSEEK_API_KEY environment variable; never hardcode it. Pricing observed on 2026-11-20: $0.30 / 1M input tokens, $1.20 / 1M output tokens, $0.006 / 1M cache read tokens.  API integration docs: https://api-docs.deepseek.com/. Billing scope and conditions: Official current model version: DeepSeek-V4.1-Flash-1017; API identifier: deepseek-flash. Peak: Monday–Friday 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays. Off-peak: all other times, including weekends and Chinese public holidays. Selected billing window: peak. Cache-hit input uses the cache-read rate; cache-miss input uses the input rate. Thinking tokens are billed as output. Context: 1M tokens; documented maximum output: 384K. Standalone Search and Fetch APIs are billed per successful request. Since November 18, 2026 at 12:00 Beijing time, model-triggered server-side search and fetch carry no separate tool fee; returned content is billed as model token usage. Token reference subtotals exclude independent API calls. Billing currency: USD. Deployment region and account availability are not established by this price page; region is unspecified, not inferred from currency or document language. Supports documented JSON output and tool calls, with thinking enabled by default and thinking/reasoning_effort controls. Control documentation: https://api-docs.deepseek.com/guides/thinking_mode/; internal evidence SHA-256: 5ab9a7b03778a41ea37046b88936a3066fcfe9085d6b7436b5fc0ec371b08a88. Output quality and strict schema adherence are not established.. Source: https://api-docs.deepseek.com/quick_start/pricing/.
Task test prompt
Read this support ticket and return a JSON object with "summary", "urgency" (low, medium, or high), and "next_step". Use null for missing details.

Ticket: I paid for the Pro plan yesterday, but my account still shows Free. My team needs access before tomorrow’s demo.
Price facts
Input: $0.30 / 1M tokens (Applies in peak window. Cache miss. Peak: Monday–Friday 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays. Off-peak: all other times, including weekends and Chinese public holidays.)
Output: $1.20 / 1M tokens (Applies in peak window. Includes thinking tokens. Peak: Monday–Friday 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays. Off-peak: all other times, including weekends and Chinese public holidays.)
Cache read: $0.006 / 1M tokens (Applies in peak window. Cache hit. Peak: Monday–Friday 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays. Off-peak: all other times, including weekends and Chinese public holidays.)
Currency: USD
Regions: unspecified
Unit source: manual
Checked: 2026-11-20
Source: https://api-docs.deepseek.com/quick_start/pricing/

Source check: 2026-11-20 · valid no later than 2026-11-23 09:20 UTC

Billing evidence and sources

deepseek/deepseek-flash · direct

Billing conditions

Official current model version: DeepSeek-V4.1-Flash-1017; API identifier: deepseek-flash. Peak: Monday–Friday 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays. Off-peak: all other times, including weekends and Chinese public holidays. Selected billing window: peak. Cache-hit input uses the cache-read rate; cache-miss input uses the input rate. Thinking tokens are billed as output. Context: 1M tokens; documented maximum output: 384K. Standalone Search and Fetch APIs are billed per successful request. Since November 18, 2026 at 12:00 Beijing time, model-triggered server-side search and fetch carry no separate tool fee; returned content is billed as model token usage. Token reference subtotals exclude independent API calls. Billing currency: USD. Deployment region and account availability are not established by this price page; region is unspecified, not inferred from currency or document language. Supports documented JSON output and tool calls, with thinking enabled by default and thinking/reasoning_effort controls. Control documentation: https://api-docs.deepseek.com/guides/thinking_mode/; internal evidence SHA-256: 5ab9a7b03778a41ea37046b88936a3066fcfe9085d6b7436b5fc0ec371b08a88. Output quality and strict schema adherence are not established.

Cache read / 1M
$0.006
Cache write / 1M
Not published
Per request
Not published
Currency
USD · Region not specified

Declared parameters: tools, response_format, thinking, reasoning_effort

Price source ↗
Compare saved channels →
All quotes, filters and usage tools
History · Billing and conditions

DeepSeek-V4.1-Flash-1017 API pricing: official and providers: 4 current published quote records across 1 exact API IDs and 1 providers. Compare API versions, currencies, regions and billing conditions separately; prices do not prove task quality or account access.

Latest recorded source check: 2026-11-20

Recorded changes

No changes recorded yet. The first successful check establishes a baseline; it does not count as a price drop or a newly launched endpoint.

Showing up to 100 recent events. Prices for unavailable endpoints are retained for reference. Failed source checks do not establish that a price stayed unchanged.

Open a quote’s billing details for cache reads/writes, request fees and endpoint instructions. Batch, free, promotional and off-peak tiers are selected separately. Context-tier, storage, search, tax and platform fees may be excluded.

Find lower-cost alternatives

Find lower-cost alternatives

Choose your existing model to find lower reference costs with matching declared technical constraints. Output quality remains unverified.

Loading verified quotes…

Users’ choices · DeepSeek-V4.1-Flash-1017

Describe your work to find related tasks.

View independent evaluation on Artificial Analysis ↗ · external link; we do not copy its scores or prices.

Classification sources

Source declarations for registered versions; not task quality, current account availability or price approval. Unknown versions do not inherit sibling classifications.

Model variants

Snapshots and routing variants remain distinct. Choose an exact ID in the comparison before deciding; model quality and access conditions need your own review.

More from this maker

DeepSeek →