Same ~60M shape as G1-nano, trained from scratch on 2× tokens (3B). Weak gains, likely a 60M ceiling. Base + instruct — lab note, not a G1 upgrade.
Jules YZERD
AZERDSQ
AI & ML interests
None yet
Recent Activity
updated a collection 4 days ago
G2 Nano updated a model about 2 months ago
AZERDSQ/G0-nano-instruct updated a model about 2 months ago
AZERDSQ/G0-nano-baseOrganizations
None yet
models 7
AZERDSQ/G0-nano-instruct
Text Generation • 62.1M • Updated • 44 • 5
AZERDSQ/G0-nano-base
Text Generation • 62.1M • Updated • 51 • 1
AZERDSQ/G2-nano-safety
Text Classification • 60M • Updated • 57 • 3
AZERDSQ/G2-nano-instruct
Text Generation • 60M • Updated • 44 • 1
AZERDSQ/G2-nano-base
Text Generation • 60M • Updated • 42 • 1
AZERDSQ/G1-nano-base
Text Generation • 60M • Updated • 14 • 1
AZERDSQ/G1-nano-instruct
Text Generation • 60M • Updated • 19 • 2
datasets 0
None public yet