DeepSeek · DeepSeek-Coder-V2

DeepSeek-Coder-V2-Instruct

READER SUMMARY

The 236B-total, 21B-active instruction checkpoint in DeepSeek's Coder-V2 family, released through weights, chat service and API access.

DOOM SCORE62.5out of 100model risk profile, not the overall index

CURRENT ASSESSMENT · REVISION 1

Why this model scores 62.5

Near-frontier coding and math performance, 128K context, API access and open commercial weights create substantial misuse and residual control challenges.

Capability71

Autonomy20

Deployment78

Misuse potential70

Control difficulty69

0 comments · 0 votesOpen discussion

No comments yet. Start the discussion.

MODEL-ATTRIBUTED EVIDENCE

News tied to DeepSeek-Coder-V2-Instruct

The model score of 62.5 rates this model's risk profile. The overall Doom Index of 63.2 measures the recalibrated complete evidence corpus. These values answer different questions.

NET MODEL-ATTRIBUTED INDEX CONTRIBUTION+0.01

Each article's leave-one-out Doom Index contribution is divided equally among the exact models named on that article. This prevents multi-model evidence from being claimed in full on several model pages. Model risk scores are recalculated from their technical profile and related news, but never feed back into the overall index.

17 Jun 2024CapabilityTOWARD

DeepSeek releases Coder-V2 models with 128K context

DeepSeek released 16B and 236B DeepSeek-Coder-V2 base and instruction checkpoints with 128K context, broad programming-language support, commercial use and an API alongside downloadable weights.

Full item contribution: +0.03
DeepSeek-Coder-V2-Instruct equal share: +0.01

Read assessment →

AUDIT TRAIL

Model score history

R1
Doom Score 62.5
New exact DeepSeek-Coder-V2 236B instruction checkpoint.
12 Aug 2026