OpenBMB

@OpenBMB

OpenBMB (Open Lab for Big Model Base) aims to build foundation models and systems towards AGI. Connect with us: discord.gg/7q3ry8Ny8K

huggingface.co/openbmb

Joined February 2022

Replies Reposts Media

Pinned
OpenBMB
@OpenBMB
Sep 7
🚀 Meet MiniCPM5-2B, a 2B-parameter language model bringing high intelligence density to the edge, now open source! It ranks #1 among open-source models under 4B parameters on the @ArtificialAnlys Intelligence Index, with a score of 23. It also scores 20 on the Agentic Index,
183
OpenBMB
@OpenBMB
17h
Same Q4_K_M quant, same tasks, same success rate, MiniCPM5-2B can get tasks done quicker. This makes local agent workloads a lot more interesting. Thanks for the comparison @atomic_chat_hq
atomic.chat
@atomic_chat_hq
18h
MiniCPM5-2B beats Qwen3.5-4B in 34% less time 👾 We gave three models the same real-world tasks and compared their attempts using Q4_K_M quants Tests: • Build a playlist within strict timing rules • Trace a checkout failure through server logs • Check stock and draft a
00:00
3
OpenBMB
@OpenBMB
20h
Very cool project. Right job for the 2B. Eight residents on Spark 1, events on the 27B, private memory per person. Would love to see the end product of the game!
Joey
@aijoey
Sep 9
Gave a 2B model a small town to live in This is Whisper Town. MiniCPM5-2B powers 8 residents on Spark 1. Qwen 3.8-27B model on Spark 2 picks town events. Each resident has their own memory. The model generates their dialogue and decides what information to share. Still a
00:00
3
OpenBMB
@OpenBMB
Sep 8
Congratulations to @OpenMAIC on this recognition at @UNESCO Digital Learning Week! 🎉 Turning PPTs and videos into interactive learning experiences is exactly the kind of practical AI application education needs. Great to see VoxCPM powering the voice customization for
OpenMAIC
@OpenMAIC
Sep 8
A huge moment for us: OpenMAIC at UNESCO Digital Learning Week. 🇺🇳 In Paris, we brought OpenMAIC into the room with educators from around the world , not just to demo AI, but to hear what it actually takes to make AI useful in their classrooms. We believe anyone should be able
5
OpenBMB
@OpenBMB
Sep 8
Thanks for testing it out and sharing these numbers! 🙌 "Run a swarm of these locally" — that's exactly the vision we had. 2B params, 1.56GB Q4, 200+ tok/s on a 4090, and still keeping up with 4B models. That's what local AI agents should look like. Can't wait to see what you
Eric ⚡️ Building...
00:00
Eric ⚡️ Building...
4