[NEW MODEL] SupraLabs just released a new model! - Supra-50M-Reasoning

In three linesSupraLabs releases Supra-50M-Reasoning, a reasoning variant of Supra-50M-Instruct. Fine-tuned on 500 synthetic samples generated by Qwen3 1.7B over 6 epochs, it produces a full thinking chain before each answer. Model and dataset are open-source.

Summary generated by Claude — human-verified