BREAKING
Thinking Machines Lab debuts Inkling
0
B
total params
0
B
active per token
0
M
context tokens
0
%
MMMU Pro
0
%
VoiceBench
0
%
SWE-Bench Verified
Two builds, one fine-tuning stack
Inkling
Full
●
Apache 2.0 terms
●
BF16 and NVFP4
●
On Hugging Face
Inkling-Small
Preview
●
~12B active params
●
Tinker fine-tuning
●
Inkling Playground
0
T
training tokens
0
TB
BF16 VRAM
0
GB
NVFP4 VRAM
Open-weights race heats up
AI NEWS BLITZ
Mira Murati's Thinking Machines Lab has released Inkling, its first open-weights model.