DeepSeek-V4-Pro

Canonical page on the main site: sinoaihub.com/models/deepseek-v4-pro

Type

model

Definition

DeepSeek-V4-Pro is a DeepSeek model in the DeepSeek-V4 family: MoE: 1.6T total / 49B active parameters; 1,048,576-token context window; 393,216 max output; open-weight; MIT license; released 2026-04-24 (V4 Preview) / 2026-08-13 (GA); status deprecated.

Key facts

Architecture: MoE: 1.6T total / 49B active parameters
Context window: 1,048,576 tokens (max output 393,216)
Weights: open; MIT license
API pricing: $0.66 input / $1.98 output per 1M tokens (USD)
Released: 2026-04-24 (V4 Preview) / 2026-08-13 (GA)
Capabilities: reasoning, coding, tool calling

Provider

deepseek

Model Family

DeepSeek-V4

Version

0813

Aliases

deepseek-v4-pro-0813

Release Date

2026-04-24 (V4 Preview) / 2026-08-13 (GA)

Release History

date	event	source	verification_date
2026-04-24	V4 Preview released — open weights, 1.6T/49B MoE, MIT license	https://www.deepseek.com/en/news/v4-preview/	2026-09-20
2026-08-13	V4-Pro GA (0813 checkpoint) shipped on the API	https://api-docs.deepseek.com/updates	2026-09-20
2026-09-10	Deprecation announced, then reversed the same day — service continues with unchanged billing	https://api-docs.deepseek.com/updates	2026-09-20

Status

deprecated

Architecture

MoE: 1.6T total / 49B active parameters

Parameter Information

total_parameters: 1.6T
active_parameters: 49B

Context Window

1048576

Maximum Output

393216

Capabilities

reasoning: Yes
coding: Yes
math: Yes
vision: No
tool_calling: Yes
function_calling: Yes
structured_output: Yes

Open Weight

Yes

License

MIT

Self Hosting

Yes

Api Available

Yes

Pricing

input_price_per_1m: 0.66
output_price_per_1m: 1.98
currency: USD
pricing_ref: deepseek

Official Api

Yes

Cloud Providers

DeepSeek Platform

Regions

unknown

Benchmark Results

benchmark	score	metric	date	source_type	source_url
HLE	42.7 (60.0 with tools)	accuracy	2026-08-13	vendor_reported	https://api-docs.deepseek.com/updates
Terminal-Bench 2.1	87.9	accuracy	2026-08-13	vendor_reported	https://api-docs.deepseek.com/updates
DeepSWE	62.7	accuracy	2026-08-13	vendor_reported	https://api-docs.deepseek.com/updates
Agents' Last Exam	25.7	accuracy	2026-08-13	vendor_reported	https://api-docs.deepseek.com/updates

Known Limitations

Deprecation announced 2026-09-10: news page says V4-Pro requests will route to V4.1-Flash after 2026-09-14 until V4.1-Pro launches, but the same-day change log says V4-Pro API service continues with unchanged billing - the official pages conflict
Vision not supported
Open-weight HF checkpoint last modified 2026-06-22; unclear whether it matches the 0813 GA checkpoint
Pricing is peak/off-peak: listed prices are off-peak; peak is 2x

Verification Status

verified

Model Versions

version	date	source	verification_date
V4 Preview	2026-04-24	https://www.deepseek.com/en/news/v4-preview/	2026-09-29
0813 (GA)	2026-08-13	https://api-docs.deepseek.com/updates	2026-09-29

License Changes

No documented license changes on record as of 2026-09-29.

Last Verified

2026-09-20

Sources

evidence_id	source_name	source_url	source_type	published	verified	confidence	conflict
src-models-deepseek-v4-pro-1	DeepSeek API docs — Models & Pricing	https://api-docs.deepseek.com/quick_start/pricing	Official documentation	—	2026-09-20	high	—
src-models-deepseek-v4-pro-2	DeepSeek API Change Log	https://api-docs.deepseek.com/updates	Official documentation	—	2026-09-20	high	Conflicts with the same-day deprecation announcement (V4-Pro → V4.1-Flash routing)
src-models-deepseek-v4-pro-3	DeepSeek-V4 Preview release	https://www.deepseek.com/en/news/v4-preview/	Official	2026-04-24	2026-09-20	high	—
src-models-deepseek-v4-pro-4	Hugging Face model card — DeepSeek-V4-Pro	https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro	Model card	—	2026-09-20	high	—