Qwen 3.8 27B matches GPT-5.6 Luna and nearly ties 1.7T-parameter DeepSeek on benchmarks
“Qwen 3.8 27B is a truly astonishing model.”
small-models
11 tracked signals on small-models.
Small open-source models can match or beat frontier performance on specific tasks at far lower cost.
“actually for specific tasks you can be at frontier or beyond frontier performance”
Qwen 3.8B runs on consumer laptops while matching frontier model performance
“if we wait a bit, we might get frontier level systems running on our laptops”
Voice agents need sub-950ms response; small models beat frontier models on latency
“A frontier model that think for a full second has already lost the room, no matter how good the answer is.”
Hugging Face shipped a working multi-agent economy running on a single 3B-parameter model.
Hugging Face released PP-OCRv6, a 50-language OCR model family ranging from 1.5M to 34.5M parameters.
Hugging Face built a multi-model finance drama simulation running five distinct AI personas on small models.
Hugging Face releases LFM2.5-2.6B, a compact model designed for local agent deployment.
Hugging Face kicks off the 'Build Small' hackathon celebrating small, fine-tunable models over large API providers.
OpenBMB promotes its compact MiniCPM models for cheaper, edge-friendly AI deployment at a 'Build Small' hackathon.
“instead of asking how can I use the largest model possible, a better question is can I build a useful project with a compact model that is better, cheaper and easier to deploy”
Cohere engineers introduce tiny open models for a 'Build Small' hackathon community session.
“Cohere is this like magical kind of company that has somehow managed to both embrace open source and consistently release open models and also build an actual business.”