Loading…
NewsroomLive
MegaTrain trains 120B-parameter models on single GPU using CPU RAMResearch · 6m agoOverFlowLight cuts urban gridlock 60% with real-time queue detectionResearch · 1h agoDeepSeek open-sources DSpark inference stack with 60–85% speedupReleases · 2h agoFive uncensored Qwen3.6-35B fine-tunes surface on HuggingFace in 24 hoursReleases · 13h agoNormGuard preserves image quality in flow-model RL fine-tuning by capping velocity inflationReleases · 14h agoPP-OCRv6 scales from 1.5M to 34.5M parameters across 50 languagesReleases · 15h ago
