DeepSeek-V4-Pro (Apr 2026)
DeepSeek-V4-Pro (Apr 2026) is an open-weights AI model from DeepSeek with an index score of 147 (GPT-4 = 100), rank 76 of 217.
147131–164Index score#76 of 217 models
$0.90/ $1.5Price per million tokensinput / output · via OpenRouter
1.05MContext windowtokens it reads at once
24 Apr 2026ReleasedReplaced by DeepSeek-V4-Pro
Where it sitson the frontier
o3 reached this score 12 months earlier.
Test resultsby area · tick = best by any model
Reasoning & knowledgeevidence
Mathsevidence
Codingevidence
Agentsevidence
Professional workno evidence yet
Novel problemsno evidence yet
Long tasksno evidence yet
Third-party results only, at the model’s best standard setting. Hover a bar for who ran it.
What others sayrank on 9 leaderboards
The Model Indexthird-party tests, our method#76of 217 · 147
LMArena · Creative writingcreative writing#40of 360 · 1446LMArena · Instruction followingfollowing detailed instructions#46of 362 · 1460LMArena · Textoverall chat quality, judged by people#52of 362 · 1464LMArena · Long promptslong, complex requests#50of 340 · 1474LMArena · Mathsmaths questions in chat#55of 346 · 1465LMArena · Codingcoding questions in chat#64of 357 · 1503Epoch Capabilities Indexoverall capability from many benchmarks#68of 231 · 150.2LMArena · WebDevbuilding web apps and front ends#42of 133 · 1581Best attop three among current models
Where to use itprices checked 18 Nov 2026
API providers 8 of 12 · per 1M tokens, in / out
Run it yourself
Official weightsdeepseek-ai/DeepSeek-V4-Pro↗Host prices from OpenRouter; going direct can differ.
Facts and sources
Released24 Apr 2026
Announcementnot linked yet
Release entryDeepSeek-V4 preview (V4-Pro/V4-Flash)
Scored asDeepSeek-V4-Pro
Output speed–
Hugging Facedeepseek-ai/DeepSeek-V4-Pro ↗