DeepSeek V4 Flash Pricing

DeepSeek V4 Flash API pricing

DeepSeek V4 Flash API pricing is $0.22 per million input tokens and $0.66 per million output tokens, verified September 14, 2026.

Input (per 1M tokens)

$0.22

Output (per 1M tokens)

$0.66

Cached input (per 1M tokens)

$0.01

Context window

1,000,000 tokens

Max output

384,000 tokens

Capabilities

DeepSeek V4 Flash supports tool use.

DeepSeek V4 Flash API pricing has changed 2 times on record, most recently on August 24, 2026 (increase).

• August 24, 2026: Increase — input $0.14 → $0.22/M (+$0.08) — output $0.66/M
• August 17, 2026: Increase — input $0.14/M — output $0.28 → $0.66/M (+$0.38)

Frequently asked questions

Short answers to the most common questions about DeepSeek V4 Flash pricing, limits, and capabilities.

Notes

Cheaper sibling of V4 Pro — surfaced by the weekly refresh agent May 28, 2026 (source: api-docs.deepseek.com). Cache-miss input $0.14, output $0.28 per MTok. [2026-05-30] Updated Context Window 128000 → 1000000 and Max Output Tokens 8192 → 384000 per DeepSeek V4 Flash spec confirmed via OpenRouter + devtk.ai. Populated Cache Read Price at $0.0028 per MTok (cache-hit) per api-docs.deepseek.com — cache-hit price is 1/50 of cache-miss input (not 1/120 as V4 Pro). Cache-hit reduction to 1/10 of launch price took effect 2026-04-26 12:15 UTC. [2026-08-17] MAJOR PRICE EVENT — DeepSeek moved V4 to a PEAK/OFF-PEAK schedule effective 2026-08-16 16:00 UTC. Output rose $0.28 → $0.66 off-peak and $1.32 peak (4.7× the old rate at peak). Peak hours 01:00–04:00 and 06:00–10:00 UTC; all other hours off-peak. Sources: engadget.com, aipricing.guru peak/off-peak update, deepseek.ai/pricing. Stored Output = the OFF-PEAK rate ($0.66) because off-peak covers 17 of 24 hours — schema has no peak/off-peak dimension. INPUT AND CACHE-HIT LEFT UNCHANGED ($0.14 / $0.0028): post-cutover input and cache rates were NOT published in any source found this run. ACTION FOR KHALED: recheck input/cache next run; consider adding peak/off-peak fields. This is a price increase with a time-of-day discount, not a cut. [2026-08-24] RESOLVED the open item from 2026-08-17 — post-cutover input and cache-hit rates are now published. Input rose $0.14 → $0.22 and Cache Read (cache-hit) rose $0.0028 → $0.007 per MTok, both OFF-PEAK, consistent with the stored off-peak convention. Peak equivalents are $0.44 input / $0.014 cache-hit / $1.32 output. Output confirmed unchanged at $0.66 off-peak. Corroborated across deepseek.ai/pricing, aipricing.guru "V4 Peak & Off-Peak", benchlm.ai and chat-deep.ai. Cache-hit is now ~1/31 of cache-miss input (was ~1/50). Peak/off-peak schema fields still not added. [2026-09-14] Verification attempt produced no clear result — recheck next run. The issue is not a price move but an apparent MODEL SUPERSESSION: sources this run date the stored rates ($0.22 input / $0.007 cache-hit / $0.66 output, off-peak) to the window 2026-08-16 through 2026-09-09 only, and describe DeepSeek V4 Flash as replaced from 2026-09-10 by DEEPSEEK V4.1 FLASH at $0.15 input / $0.003 cache-hit / $0.60 output off-peak (peak $0.30 / $0.006 / $1.20). V4.1 Flash is quoted with a 1M context window and 384K max output, i.e. the same spec as this row. All prices LEFT UNCHANGED — repricing this row to V4.1 rates would silently rename the SKU, and per the task spec new models are not added automatically. ACTION FOR KHALED — pick one: (a) reprice and rename this row to V4.1 Flash, or (b) set this row to Deprecated and add V4.1 Flash as a new row. Until then the calculator is quoting a tier that may no longer be purchasable. Sources: deepseek.ai/pricing, devtk.ai V4 Flash model page, opslyft, benchlm.ai.