Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
cosmoem-0.1-4x1b	24.4	50.71	37.16	29.09	35.34

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
cosmo-1b	22.97	52.01	38.02	28.73	35.43

Overall decrease in capability. Data mostly resembles what it already saw; done in hopes of experts 'settling in'.

Safetensors

Model size

5.37B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Lambent/cosmoem-0.1-4x1b

Quantizations