Context News

YES

uses some non-AdamW optimizer

Some image input (multimodality)

>= 50% on TerminalBench (https://www.tbench.ai/leaderboard)

>= 60% on BrowseComp (https://www.kaggle.com/benchmarks/openai/browsecomp)

1M+ Context

>=25T pretraining tokens

>= 73% on SWE-Bench Verified (according to epoch.ai)

>= 2 shared experts

Some variation of NSA (Native Sparse Attention)

>=16 active experts

Gemini 2.5 Pro tier or higher on FictionBench (90.6%+ at 192k)

>=512 experts

intra-expert communication

DS-MoE with adaptative expert count

>= 44% on Humanity's Last Exam (text only) at scale.com leaderboard

DeepSeek reports some results with a full-blown deep research agent, and emphasizes that this is the intended use-mode

>=52B active parameters

>=1.5T parameters

Other

2042

2043

2044

2045

2041

2040

2027

2039

2037

2034

2035

2036

2038

2024

2025

2026

2028

2029

2030

2031

2032

NO

NO

YES

At least 1.1 Million

At least 2 million

At least 4 million

At least 10 million

Other

NO

YES