DeepSeek V4
DeepSeek · released 24 Apr 2026 · open weights
- Open weights, self-hostable
Superseded by DeepSeek V4.1 Flash on 2026-09-10, and being phased out: DeepSeek routes deepseek-v4-pro requests to Flash from 2026-09-14 until a V4.1 Pro launches. Open-weights MoE (~32–37B active) with sparse attention.
Context window
1M
1,000,000 tokens
Input price
$1.32
per 1M tokens
Output price
$3.96
per 1M tokens
Blended
$1.98
3:1 in:out mix
Weights
Open
self-hostable
Released
24 Apr 2026
DeepSeek
Long-context / tier pricing — deepseek-v4-pro, peak rate, cache-miss input. Off-peak is exactly half ($0.66/$1.98) and covers every hour outside 01:00-04:00 and 06:00-10:00 UTC on weekdays. Cache hits are $0.044 peak. From 04:00 UTC on 2026-09-14 this model name is served by V4.1 Flash and billed at Flash rates. Listed prices are the base rate for prompts up to ~200K tokens.
Open-weights API prices are representative third-party-hosted rates; self-hosting costs only your own hardware.
Where it ranks
Of 40 models tracked.
Cost to run
#17
of 40
Input price
#18
of 40
Context window
#9(tied)
of 40
Compare with
The nearest models by blended cost to run.
Put it to work
Specs from BitByteCore's hand-verified Frontier Ledger (v2026-09-26), this row verified 10 Sep 2026 against DeepSeek's official pricing ↗. Model pricing moves fast, so confirm with DeepSeek before you rely on a figure.