{"acceptedBy":{"actor":"ai","name":null,"on":"2026-10-22"},"aliases":[],"change":{"last4":0,"percent":null,"prior4":0},"description":{"editedBy":null,"state":"none"},"href":"/wiki/concept-vram/","itemCount":4,"key":"concept--vram","kind":"concept","name":"VRAM","neighbours":[{"key":"licence--apache-2-0","kind":"licence","name":"Apache-2.0","relation":"mentioned_with","weight":3},{"key":"platform--hugging-face","kind":"platform","name":"Hugging Face","relation":"mentioned_with","weight":3},{"key":"concept--real-time-factor","kind":"concept","name":"real-time factor","relation":"mentioned_with","weight":2},{"key":"maker--mistral-ai","kind":"maker","name":"Mistral AI","relation":"mentioned_with","weight":2},{"key":"task--voice-cloning","kind":"task","name":"voice cloning","relation":"mentioned_with","weight":2},{"key":"licence--lfm-1-0","kind":"licence","name":"LFM 1.0","relation":"mentioned_with","weight":1},{"key":"maker--nineninesix","kind":"maker","name":"NineNineSix","relation":"mentioned_with","weight":1},{"key":"maker--openbmb","kind":"maker","name":"OpenBMB","relation":"mentioned_with","weight":1},{"key":"model--gpt-oss-safeguard-20b","kind":"model","name":"GPT-OSS-Safeguard-20B","relation":"mentioned_with","weight":1},{"key":"model--kanitts2-en","kind":"model","name":"KaniTTS2-en","relation":"mentioned_with","weight":1},{"key":"model--llavaguard","kind":"model","name":"LlavaGuard","relation":"mentioned_with","weight":1},{"key":"model--llavaguard-7b","kind":"model","name":"LlavaGuard-7B","relation":"mentioned_with","weight":1}],"relatedItems":[{"publishedAt":"2026-07-16T00:00:00Z","reason":"concept--vram","title":"Mistral AI releases Shieldstral 1.0, a 3B multimodal safety classifier under Apache 2.0","url":"/decisions/mistral-shieldstral-1-0-3b/"},{"publishedAt":"2026-04-03T00:00:00Z","reason":"concept--vram","title":"OpenBMB releases VoxCPM2, a 2B Apache 2.0 TTS with voice design and 48 kHz output","url":"/open-weights/openbmb-voxcpm2/"},{"publishedAt":"2026-02-12T00:00:00Z","reason":"concept--vram","title":"NineNineSix releases KaniTTS2-en, a 400M English TTS with voice cloning","url":"/open-weights/nineninesix-kani-tts-2-en/"},{"publishedAt":"2025-10-31T00:00:00Z","reason":"concept--vram","title":"Mistral releases Ministral 3, edge models in 3B, 8B and 14B under Apache 2.0","url":"/open-weights/mistralai-ministral-3/"}],"schemaVersion":1,"series26":[0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0],"series8":[0,0,0,0,0,0,0,0],"weeksOfHistory":52}