Soofi S releases a 30B-A3B open Mamba-MoE model trained on 27T tokens
_ScottCondron · x · 2026-07-23
The post announces the Soofi S pretraining tech report, describing a 30B-A3B mixture-of-experts hybrid Mamba model trained on about 27T tokens with intentionally up-weighted German.
Key claims from the report:
- It is the strongest fully open model in the authors’ evaluations on both the English and German aggregates, ahead of Olmo 3 32B and Apertus 70B.
- The release includes unusually complete transparency: per-source data accounting, hyperparameters, training and eval code, checkpoints, and a project page.
- The project was trained on Deutsche Telekom infrastructure, and the authors even released a Weights & Biases dashboard for the run.
The post is basically celebrating how unusually open the whole training stack is, not just the model weights.
Related event: Soofi S Releases 30B Hybrid Mamba Open-Source Model(2 posts)→
More from Models
- Gemini 3.5 Flash-Lite is 71x Cheaper Than Claude for Doc Extraction — rseroter · 2026-07-23
- Sentdex questions SemiAnalysis’ chart comparing Kimi K3 with Nemotron 3 Ultra — Daniel_Farinax · 2026-07-23
- A VRAM-by-VRAM model roundup maps the best open weights from 4GB to 384GB — victormustar · 2026-07-23
- BTL-3 launches as a 27B open-weight agentic coding model with 95.12% HumanEval — soteko · 2026-07-23
- Testing LLMs with a Churchill Insult Prompt: GPT Outperforms Kimi and Gemini — emollick · 2026-07-23
- NVIDIA Recaps Kaggle Nemotron Challenge: 5,000+ Explore Open Model Reasoning — NVIDIAAI · 2026-07-23