API PRICES

DeepSeek · deepseek-v4-flash

Provider prices, cheapest first

Browse all models

DeepSeek · deepseek-v4-flash official price: not yet recorded. Cheapest inference provider: DeepInfra, input 0.09, output 0.18 USD per million tokens. Relays have only older quotes; the lowest is ai.38yun.cn, group "GPT stable channel [machine translation] (on site: GPT 稳定渠道)" 0 USD_CREDIT per million total tokens (credit price, not cash, observed 2026-10-10). 224 sites have comparable prices; 528 sites list this model.Relay blended price uses 75% input and 25% output.

Official prices are the vendor's cash prices. Inference providers also list cash prices and can be compared directly. Relays list prices in site credit; what you actually pay depends on each site's top-up rate, so they cannot be compared with official prices directly.

Inference providers

ProviderInputOutputCache readCache writeUnitObservedConditions and sourcevs official
DeepInfra0.090.180.018—USD / 1M tokens
Official source
OpenRouter · Aggregator0.0910.1820.0182—USD / 1M tokens
Official source
SiliconFlow3.009.000.30—CNY / 1M tokens
Official source

Unit: USD credits per 1M tokens. These are site-credit prices; actual cost depends on each site’s top-up rate.

Searching…

Browse all providers

Blended price = input × 75% + output × 25%, without cache.

sites have comparable prices224528 sites list this model; approximately 131 operators
Older lowest input price (reference)0.03575USD credits / 1M
Median of each provider’s lowest price0.695USD credits / 1M

Relays

Prices below are in site credit, not cash.

#ProviderPlan / conditionsInput ↑USD credits / 1MOutput USD credits / 1MCache readUSD credits / 1MBlendedUSD credits / 1MDetails
Older price
26sheapi.topAsk or discussCollected:
View 1 plans
deepseek
deepseek series models, in-house special-supply channel [machine translation]Original: deepseek系列模型,自研特供渠道
0.210.630.006999999999930.315Full prices and sources
27ai.lt4net.orgAsk or discussCollected:
View 1 plans
opencode0.240.760000000000080.007999999999920.37000000000002Full prices and sources
27gptapi.asiaAsk or discussCollected:
View 2 plans
2 more sites of the same company

Basis: identical homepage assets, identical quotes, same status-page name, logo or server address. Counted as one company in the ranking; this is not a legal finding.

default
Default channel [machine translation]Original: 默认通道
0.240.96—0.42Full prices and sources
29realmrouter.cnAsk or discussCollected:
View 1 plans
default
Default [machine translation]Original: 默认
0.280.560.00560.35Full prices and sources
30aiapi.worldAsk or discussCollected:
View 1 plans
cmodels
cmodels
0.2890.867—0.4335Full prices and sources
31llm.azopenai.comAsk or discussCollected:
View 1 plans
default
default
0.31.20.0060.525Full prices and sources
32tokenspace.cloudAsk or discussCollected:
View 2 plans
default_special_20.3361.0080.0111999999998880.504Full prices and sources
33mumuqiu.cyouAsk or discussCollected:
View 2 plans
Distribution [machine translation]Group on site: 分发
Group only for the site owner to hand out perks to chat group members, do not use!!! [machine translation]Original: 仅用于站长给群友发福利的分组,不要用!!!
0.350.70.007999999999950.4375Full prices and sources
34api.hubagi.aiAsk or discussCollected:
View 2 plans
Special supply [machine translation]Group on site: 特供
Special supply [machine translation]Original: 特供
0.3751.125—0.5625Full prices and sources
35555codex.xyzAsk or discussCollected:
View 1 plans
Chinese models [machine translation]Group on site: 国产模型
Chinese models [machine translation]Original: 国产模型
0.391.170.012999999999870.585Full prices and sources
36pro.auroai.icuAsk or discussCollected:
View 1 plans
Chinese [machine translation]Group on site: 国
Chinese [machine translation]Original: 国
0.41.60.010.7Full prices and sources
37api.apiyi.comAsk or discussCollected:
View 4 plans
1 more site of the same company

Basis: identical homepage assets, identical quotes, same status-page name, logo or server address. Counted as one company in the ranking; this is not a legal finding.

ClaudeCode
ClaudeCode
0.4181.2540.0133760.627Full prices and sources
38lanyiapi.comAsk or discussCollected:
View 1 plans
2 more sites of the same company

Basis: identical homepage assets. Counted as one company in the ranking; this is not a legal finding.

zh-model
Chinese models [machine translation]Original: 国产模型
0.441.320.013999999999920.66Full prices and sources
38api.gravitex.aiAsk or discussCollected:
View 1 plans
default
Default group [machine translation]Original: 默认分组
0.441.329999999999880.030000000000080.66249999999997Full prices and sources
40mzhcloud.comAsk or discussCollected:
View 2 plans
Chinese models [machine translation]Group on site: 国产模型
ds low cache, includes deepseek, glm, kimi, mimo, hunyuan, qwen, Minimax, etc. [machine translation]Original: ds低缓存,包括deepseek、glm、kimi、mimo、hunyuan、qwen、Minimax等
0.451.350.014999999999850.675Full prices and sources
41bikip.proAsk or discussCollected:
View 1 plans
deepseek-20.4841.4520.0153999999999120.726Full prices and sources
42api.getezclaw.comAsk or discussCollected:
View 1 plans
DeepSeek0.490.980.0490.6125Full prices and sources
431yapi.comAsk or discussCollected:
View 1 plans
Chinese models [machine translation]Group on site: 国产模型
Chinese models · GLM/DeepSeek/Kimi/MiniMax/Qwen [machine translation]Original: 国产模型 · GLM/DeepSeek/Kimi/MiniMax/Qwen
0.50.5—0.5Full prices and sources
43love.qinyan.icuAsk or discussCollected:
View 2 plans
1 more site of the same company

Basis: identical quotes, same status-page name, logo or server address. Counted as one company in the ranking; this is not a legal finding.

Official full-strength direct connection deep, Grok all models (grok400K context, image recognition, ultra-low rate multiplier) [machine translation]Group on site: 官满直连deep、Grok全模型(grok400K上下文可识图 倍率超低)
Official relay direct full-strength 🐋Deepseek all models, 1000K context, image recognition (compatible with CC and Gemini formats) ¥0.4/dao (US dollar) [machine translation]Original: 官转直连满血🐋Deepseek全模型1000K上下文可识图(适配CC和Gemini格式) ¥0.4/刀
0.50.50.050.5Full prices and sources
43api.ai-gaochao.cnAsk or discussCollected:
View 18 plans
guo_chan
guo_chan
0.510.10.625Full prices and sources
43code.aifast.topAsk or discussCollected:
View 1 plans
Chinese models [machine translation]Group on site: 国产模型
Self-deployed + official relay deepseek/kimi/glm/minimax Chinese models, half the official price [machine translation]Original: 自部署+官转 deepseek/kimi/glm/minimax国产模型,官方半价
0.510.10.625Full prices and sources
43codexcn.topAsk or discussCollected:
View 1 plans
default0.510.050.625Full prices and sources
48123ai.ai-wx.cnAsk or discussCollected:
View 2 plans
1 more site of the same company

Basis: identical quotes. Counted as one company in the ranking; this is not a legal finding.

gc-chat
Chinese text group [machine translation]Original: 国产文字分组
0.61.20.120.75Full prices and sources
48api.mooko.aiAsk or discussCollected:
View 1 plans
DeepSeek dedicated line [machine translation]Group on site: DeepSeek专线
DeepSeek official direct connection dedicated line [machine translation]Original: DeepSeek官方直连专线
0.61.20.0120.75Full prices and sources
48api.zygtoken.comAsk or discussCollected:
View 1 plans
Chinese models [machine translation]Group on site: 国产模型 0.61.20.0120.75Full prices and sources
Special groups · 17 rows · not ranked

These plans are marked as test, internal, disabled, gated by top-up or membership, or have a near-zero ratio. They may not be purchasable or stable, so they are listed separately and left out of the ranking and statistics.

  • test or debug group: 11 rows
  • zero-ratio group: 4 rows
  • marked display-only or do-not-select: 3 rows
  • near-zero-ratio group: 1 rows
  • ai.38yun.cn · Professional free model channel [machine translation]Group on site: 专业免费模型通道 · deepseek-ai/deepseek-v4-flash · Input 0 / Output 0 USD credits / 1M
  • api.wapq.cn · Perk [machine translation]Group on site: 福利 · deepseek-v4-flash · Input 0 / Output 0 USD credits / 1M
  • omnigate.cc · deepseek-v4-flash free channel, hammer it [machine translation]Group on site: deepseek-v4-flash免费渠道,请猛攻 · deepseek-v4-flash · Input 0 / Output 0 USD credits / 1M
  • wqdsjhaoai.site · amd free models [machine translation]Group on site: amd免费模型 · DeepSeek-V4-Flash · Input 0 / Output 0 USD credits / 1M
  • 222650156.xyz · Perk free [machine translation]Group on site: 福利free · free/deepseek-v4-flash · Input 0.0001 / Output 0.0002 USD credits / 1M
  • ai.daoyivip.com · Chinese models test [machine translation]Group on site: 国模测试 · deepseek-v4-flash · Input 0.1 / Output 0.2 USD credits / 1M
  • realmrouter.cn · Chinese models - test [machine translation]Group on site: 国模-test · deepseek-v4-flash · Input 0.112 / Output 0.224 USD credits / 1M
  • hk.apitoken.ai · sandbox-China · deepseek-v4-flash · Input 0.14 / Output 0.28 USD credits / 1M
  • hk.apitoken.ai · sandbox · DeepSeek-V4-Flash · Input 0.14 / Output 0.28 USD credits / 1M
  • hk.apitoken.ai · sandbox-China · DeepSeek-V4-Flash · Input 0.14 / Output 0.28 USD credits / 1M
View all 17 rows →
Related variants · 10 rows · not ranked

These are special editions of the model (thinking, max, search, batch, long context, free, etc.). They are priced differently from the base model, so they are listed separately and not ranked. Names that only add an effort level such as low, high or nothinking are compared with the base model.

  • deepseek-v4-flash-free (-free · free variant): 3 rows
  • deepseek-v4-flash-max (-max · max edition): 3 rows
  • deepseek-v4-flash-thinking (-thinking · thinking edition): 2 rows
  • deepseek-v4-flash:free (:free · free variant): 1 rows
  • zen/deepseek-v4-flash-free (-free · free variant): 1 rows
  • ai.bytonsec.com · zen · zen/deepseek-v4-flash-free · Input 0 / Output 0 USD credits / 1M
  • ai.bytonsec.com · zen · deepseek-v4-flash-free · Input 0 / Output 0 USD credits / 1M
  • xckjj.com · Chinese models [machine translation]Group on site: 国模 · deepseek-v4-flash:free · Input 0.0112 / Output 0.0224 USD credits / 1M
  • nty-hk.zhouyang168.top · free · deepseek-v4-flash-free · Input 0.1 / Output 0.1 USD credits / 1M
  • yfy.zhouyang168.top · free · deepseek-v4-flash-free · Input 0.1 / Output 0.1 USD credits / 1M
  • api.pezayo.com · default · deepseek-v4-flash-thinking · Input 0.143 / Output 0.286 USD credits / 1M
  • api.pezayo.com · official · deepseek-v4-flash-thinking · Input 0.143 / Output 0.286 USD credits / 1M
  • yinli.one · deepseek · deepseek-v4-flash-max · Input 1 / Output 2 USD credits / 1M
  • apinexus.dpdns.org · default · deepseek-v4-flash-max · Input 60 / Output 180 USD credits / 1M
  • apinexus.dpdns.org · vip · deepseek-v4-flash-max · Input 90 / Output 270 USD credits / 1M
View all 10 rows →
How this ranking works

All per-token prices: 96,427 prices from 839 sites (about 733 companies). 88,950 are ranked, from 826 sites (about 720 companies); 7,477 more are listed separately and not ranked. Prices collected up to 2026-10-10.

This model has 1061 recorded prices; 400 can be compared directly (366 ranked, 34 listed separately). Not comparable yet: tiered by rules (see Tiered prices) 291, price not parsed yet 133, new price under review; showing the previous one 100, billed per request or another unit 63, display unit not confirmed 42, ratio looks unset 24, scope of the official doc price not confirmed 8

0 recent · 284 stale (older than 7 days, or not updated by this refresh) · 0 unverified timestamps. Only older prices are available, so the statistics are for reference only. 3 zero-valued pairs are separately unverified, without proof of free eligibility. Timing does not verify payment, upstream identity or quality.

Prices are converted to site credits using each site’s published ratio. What you actually pay depends on its top-up rate. Input, output and cache prices in each row come from the same plan. Several sites run by the same company take one place in the ranking: its cheapest plan, with the other sites listed under it. We judge “same company” by identical prices, status pages or homepage code; this is not a legal finding. To see every site’s prices, choose “All plans” or filter by provider.

“Self-declared unofficial channel” means the plan name itself says “reverse-engineered” or similar. We have not verified their source or quality. Minimum and median only count regular plans.

Full methodology →

Switch model