BREAKING
Thinking Machines Releases Inkling
0
B
total params
0
B
active params
0
M
context tokens
0
T
pretrain tokens
0
%
VoiceBench
0
%
MMAU
Controllable Thinking Effort
Trained on NVIDIA GB300 NVL72
GB300 NVL72
hardware
●
72 Blackwell Ultra GPUs
●
36 Grace CPUs
●
130 TB/s NVLink
Open stack
deploy
●
NVFP4 checkpoint on Hugging Face
●
SGLang serving support
●
Fine-tuning via Tinker
A notable open multimodal entry
AI NEWS BLITZ
Thinking Machines Lab has launched Inkling, its first open-weight foundation model.