DeepSeek V4 Pro

Name: DeepSeek V4 Pro model facts
Creator: OSMEXO

deepseek/deepseek-v4-pro

DeepSeek
Family · DeepSeek V4 Pro
Released 13 Aug 2026
Open weights
Text → Text
Reasoning · adjustable
Tools · JSON

Live on OSMEXO

Try now Request a discount

API setup & price →

Spec & list price

Context1M

Max output384K

Input / 1M$1.60

Output / 1M$3.202× input

Scores · standing

Intelligence36.0#48/489

Coding68.8#29/196

Usage

Tokens / day64.4B▼ 12% 7d

Usage rank#42

Requests / day1.4M

Tool calls / req0.28

Per request · 30 days

Avg input24.4K1.6× median

Avg output8801.3× median

Cache hit83%1.2× median

Reasoning share60%about median

Tool-call failures2.3%1.2× median

Consumption

Peak 119.6B 5 OctLow 28.7B 31 OctRank today #42

News & trends

10 mentions this week ▲ 11%this model · videos, posts, articles · 30 days

Trending · 14 days

News · latest from DeepSeek

Tell HN: Mistrap Pro includes DeepSeek v4.1 Flash 1017docs.mistral.ai · 1 Nov
Show HN: DeepSeek v4.1 Flash-1017 Safety Evaluationtwitter.com · 26 Oct
DeepSeek V4 Pro 0813 vs Claude Fable 5 on DeepSWE: Cost, Coding, and RoutingTogether AI · 17 Aug
DeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and CodingTogether AI · 6 Aug

Benchmarks

bar = share of the top score · Benchmark data: Artificial Analysis

GPQA Diamond#15 of 445

92.8%

Long-context reasoning#46 of 393

80.3%

Terminal-Bench 2.1#25 of 182

78.7%

Coding index#29 of 196

68.8

τ-Bench Banking#18 of 154

39.6%

SciCode#54 of 160

51.0%

Humanity's Last Exam#46 of 465

41.0%

Intelligence index#48 of 489

36.0

Speed · median across 15 providers, latest day

Output speed#237/33348 tok/s1.4× slower than median

Fastest provider88 tok/sbest of 15 providers

First token#1991.36 s1.3× longer than median

First answer token#7526 sthinking included

#8 cheapest of 15 with a similar score

Value

Intelligence Index

Blended price per 1M tokens (log)

Family progress

Intelligence Index

Family releases36.0

Technical details

Reasoning: on · high
Tokenizer: DeepSeek
License: MIT
Parameters: 1.6T
Moderation: none
Downloads · likes: 369.3K · 5.7K

Reasoning effort

xhigh
high · default
none

Request parameters 21

tools
tool_choice
response_format
structured_outputs
reasoning
reasoning_effort
include_reasoning
temperature
top_p
top_k

+11 more

min_p
max_tokens
max_completion_tokens
stop
seed
frequency_penalty
logit_bias
logprobs
presence_penalty
repetition_penalty
top_logprobs

Interesting facts

6× the output room of similar models (393.2K against a 65.5K median).
Released 80 days ago, the newest of 2 DeepSeek V4 Pro models.

When to use

Best for

Difficult science and reasoning questions
Coding agents working across large codebases
Long document analysis and Q&A
Self-hosted deployments (open weights)

Think twice for

Image or screenshot input (text only)
High-volume simple chat (a Flash tier costs less)
Latency-sensitive interactive features

Research papers

Same family