google/gemma-4-E2B-it

Any-to-Any • 5B • Updated Jul 20 • 2.94M • 1.07k

Note Best overall for day-to-day use

LiquidAI/LFM2.5-2.6B

Text Generation • 3B • Updated about 15 hours ago • 99.8k • 852

Note Great for agentic coding

Qwen/Qwen3.5-2B-Base

Image-Text-to-Text • 2B • Updated Apr 23 • 303k • 105

Note Take care to use the Base version to avoid doom loops

openbmb/MiniCPM5-2B

Text Generation • 3B • Updated 25 days ago • 1.25M • 1.81k

Note Unusable for agentic coding, and seems overfitted to benchmarks