Open-Source Base Models Expected to Double in Size to Keep Pace

The GLM base model is anticipated to double in size for its 5.5 version, while DSv4 is expected to maintain its current size with room for post-training growth. Kimi currently leads open-source models at 2.8 trillion parameters, with Longcat 2 at 1.8 trillion parameters also noted.

So What

The increasing scale of open-source models will drive higher compute and memory demand for training and inference, impacting infrastructure providers.

View Tweet→

Sources (1)

Open-Source Base Models Expected to Double in Size to Keep Pace

Semicon Twitter · Twitter · 5d ago

View tweet→

Credibility

Lead sourceSemicon Twitter

Corroboration1 distinct outlet

VerificationSingle source

Unlock the full signal with Pro

Tessara synthesizes 80+ data sources into actionable intelligence for AI infrastructure investors.

Unlock with Pro→Sign in

← SignalsSignal

Notable5.7ModelsSingle sourceactive

Updated 5d ago · 1 source

Open-Source Base Models Expected to Double in Size to Keep Pace

The GLM base model is anticipated to double in size for its 5.5 version, while DSv4 is expected to maintain its current size with room for post-training growth. Kimi currently leads open-source models at 2.8 trillion parameters, with Longcat 2 at 1.8 trillion parameters also noted.

So What

The increasing scale of open-source models will drive higher compute and memory demand for training and inference, impacting infrastructure providers.

View Tweet→