Agent Mode News

Cursor (custom model)

Anthropic

Google

OpenAI

Other

DeepSeek

Anthropic

OpenAI

Google

xAI

Meta

DeepSeek

Other

NO

Some variation of NSA (Native Sparse Attention)

>=25T pretraining tokens

>= 2 shared experts

>=52B active parameters

>=16 active experts

uses some non-AdamW optimizer

>=512 experts

1M+ Context

intra-expert communication

Some image input (multimodality)

>=1.5T parameters

DS-MoE with adaptative expert count

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

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

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

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

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

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

NO

YES

YES