Atlas

Models

← All models

DeepSeek-R1-0528

DeepSeek · 2025-05-28 · 684.5B parameters

DeepSeek-R1-0528 is the May 28, 2025 minor-version update to DeepSeek-R1, retaining its DeepSeek-V3 MoE backbone (reported as 671B main-model parameters with 37B activated per token; the released checkpoint also contains one multi-token-prediction module). DeepSeek attributes stronger reasoning to increased post-training compute and algorithmic optimization; on AIME, average reasoning length rose from 12K to 23K tokens. The release also reduces hallucinations and improves system-prompt, JSON-output, function-calling, and coding support.

Benchmark scores

BenchmarkScore
AA-LCR54.7
AA-Omniscience Index-27.1
Aider Polyglot71.6
AIME 202589.2
ALE-Bench1059
Apex (MathArena)1.0
ARC-AGI-21.1
Arena Text — Business, Management & Financial Ops (No Style Control)1399
Arena Text — Chinese (No Style Control)1460
Arena Text — Creative Writing (No Style Control)1410
Arena Text — English (No Style Control)1435
Arena Text — Entertainment, Sports & Media (No Style Control)1402
Arena Text — Exclude Ties (No Style Control)1420
Arena Text — Exclude Ties (Style Controlled)1413
Arena Text — Expert (No Style Control)1397
Arena Text — French (No Style Control)1420
Arena Text — French (Style Controlled)1422
Arena Text — German (No Style Control)1398
Arena Text — Hard Prompts (No Style Control)1416
Arena Text — Hard Prompts English (No Style Control)1418
Arena Text — Instruction Following (No Style Control)1382
Arena Text — Japanese (No Style Control)1384
Arena Text — Korean (No Style Control)1355
Arena Text — Legal & Government (No Style Control)1420
Arena Text — Life, Physical & Social Science (No Style Control)1431
Arena Text — Longer Query (No Style Control)1392
Arena Text — Medicine & Healthcare (No Style Control)1447
Arena Text — Multi-turn (No Style Control)1405
Arena Text — Non-English (No Style Control)1415
Arena Text — Non-English (Style Controlled)1406
Loading Atlas data…