DeepSeek R&D Coding Benchmark

DeepSeek-V4-Pro 67% Only in system cards

Coding agents complete 30 tasks taken from DeepSeek internal R&D work in PyTorch, CUDA, Rust and C++. Human-written rubrics score each task.

In-house DeepSeek evaluation. The tasks come from internal repositories and are not public.

Made by DeepSeek Released Apr 2026 Fixing real repositories Tasks kept private Last in a system card Apr 2026 Last model added Apr 2026

Level: Expert Specialists with years of training, often over hours of work
Scored: Graded by people Human experts grade or compare the outputs.
Answers: Agent task Many steps in a terminal, browser, desktop, app, game or robot.

Model scored in lab cards

System card reports it, from 1 lab

DeepSeek-V4-Pro leads with 67%

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

Bars Table

67.0 DeepSeek-V4-Pro

1 DeepSeek-V4-Pro 67%

#	Model	Score	Setting	Date	Source
1	DeepSeek-V4-Pro	67%		Apr 2026	Model card

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card

Labs report it only in system cards

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

Date	Document	Score	As the document writes it	Section
26 Apr 2026	DeepSeek-V4-Pro technical report	67	R&D Coding Benchmark, Pass Rate, DeepSeek-V4-Pro-Max, 30 internal tasks (Table 8)	Capabilities

DeepSeek-V4-Pro leads with 67%

Scores over time

Labs report it only in system cards

One system card reports it

More in Coding