DeepSeek V4 Pro (latest)

deepseek/deepseek-v4-pro

hybrid-attention long-context reasoning.

cmd --model deepseek/deepseek-v4-pro
Intelligence index
36.3
Output speed
65.3 tok/s
Input
$0.66 /M
Output
$1.98 /M
Cache read
$0.02 /M
Agent-loop cost
$0.21 /M in
Context window
1M tokens
Released
August 13, 2026
Modalities

vs. the lineup

DeepSeek V4 Pro (latest) beside its stablemates and nearest rivals. The ◆ marks the best value in each column across every row shown.

pin a rival:
ModelIntelligenceCodingSpeedInput $/MOutput $/MBlended $/MContext
Muse Spark 1.3 Contributor48.275.8418.4$0.10$0.20$0.131.05M
DeepSeek V4.1 Flash39.5246.8$0.15$0.60$0.261M
DeepSeek V4 Pro (latest)36.368.865.3$0.66$1.98$0.991M
DeepSeek V4 Flash Vision (exp)3565228.9$0.15$0.60$0.261M
DeepSeek V4 Flash (latest)3569.1236.4$0.15$0.60$0.261M
DeepSeek V4 Flash Fast$0.28$0.56$0.351M
Kimi K3pinned43.876.237.2$3$15$61M
Kimi K2.7 Codepinned26.360.842.5$0.95$4$1.71256K

coding performance

The Intelligence Index and its sub-scores, ranked against every scored model in the catalog. A metric that has not been measured for DeepSeek V4 Pro (latest) has been left empty.

Coding Index
68.8
#30 of 53 scored
Terminal-Bench
78.7
#25 of 53 scored
Intelligence Index
36.3
#26 of 60 scored
Long-context reasoning
80.3
reasoning across a long context
SciCode
51
scientific coding
GPQA Diamond
92.8
graduate-level QA

peak hours

DeepSeek V4 Pro (latest) is not billed at one flat rate — the price changes with the time of day. Every other price on this page is the off-peak rate, which is what a request costs 17 hours out of 24. Peak runs Monday to Friday only, so it is never charged at the weekend.

BandHours (UTC)Input $/MOutput $/MCache read $/M
Off-peak17h/day$0.66$1.98$0.02
Peak7h/day · 01–04 & 06–10 UTC, Mon–Fri$1.32$3.96$0.04

usage calculator

How far a month of credits goes on DeepSeek V4 Pro (latest).

Input tokensfresh prompt
800
Output tokensmodel reply
180
Cache read tokensre-read context
50K
cost / request $0.0020 · in $0.66 · out $1.98 · cache $0.02 per M
fresh input 27%output 18%cache reads 55%
Requests / 30 days
5.0K
$10 credits ÷ $0.0020 per request
~1.0K quick fixes~202 bug fixes~34 feature PRs

what real work costs

Real coding tasks priced end to end on DeepSeek V4 Pro (latest), from a quick lookup to a full-repo agent run.

One agent task
$0.06
180K in at 75% cache hit, 12K out
What you pay
$0.29 /M
all-in across every token that task touched
Sticker input
$0.66 /M
cache reads bill at $0.02 /M instead
TaskTokens in · outDeepSeek V4 Pro (latest)Muse Spark 1.3 ContributorClaude Haiku 4.5
Quick lookup / one-liner8K · 1K$0.0027$0.0003$0.0065
Review a 500-line PR60K · 4K$0.02$0.0027$0.04
Fix a bug (agent loop)180K · 12K$0.06$0.0072$0.12
Refactor a module320K · 20K$0.10$0.01$0.21
Full-repo agent run900K · 45K$0.24$0.03$0.49

frequently asked

How much does DeepSeek V4 Pro (latest) cost?
$0.66/M input and $1.98/M output, cache reads $0.02/M. In an agent loop most input is cache-read, so the effective input rate is about $0.21/M.
Which plan do I need?
Available on Go and above.
How do I switch to it?
Run cmd --model deepseek/deepseek-v4-pro, or type /model in a session and pick it. You can switch mid-session without losing context.

Ship code that matches your taste

Command Code is the AI coding agent that continuously learns your taste. Start for $1.

Benchmarks from Artificial Analysis (v4.3)commandcode.ai/models/deepseek-v4-pro