All AI models compared: benchmarks and prices

Fits in, GB

Model	Humanity's Last Exam	Per task	Price	Speed	Context
Granite 4.2 30BIBMOpen	11.2thinking	$0.072	$0.16 / $0.65	48	128K
Granite 4.2 8BIBMOpen	9.7thinking	$0.024	$0.06 / $0.25	65	128K
Granite 4.2 3BIBMOpen	6.6thinking	$0.006	$0.03 / $0.12	56	128K
Granite 4.0 H 350MIBMOpen	6.4	—		—	33K
Granite 4.0 350MIBMOpen	5.5non-thinking	—		—	33K
Granite 4.0 MicroIBMOpen	5non-thinking	—	$0.017 / $0.11	—	131K
Granite 4.0 H 1BIBMOpen	5	—		—	128K
Granite 4.0 1BIBMOpen	4.8non-thinking	—		—	128K
Granite 4.1 30BIBMOpen	4.1	—	$0.10 / $0.25	—	131K
Granite 4.0 H SmallIBMOpen	3.8non-thinking	—	$0.05 / $0.22	—	128K
Granite 4.1 8BIBMOpen	3.8non-thinking	—		—	131K
Granite 4.1 3BIBMOpen	3.4	—		—	131K