Four open-weight models in 72 hours — only NVIDIA's fits in 24GB VRAM
BTA_Labs · reddit · 2026-10-19
A Reddit recap of four open-weight model drops in 72 hours, charted against the 24GB VRAM line:
- Meta Muse Spark Open: 816B total/42B active/1M context. Its 'desktop' demo is two 512GB Mac Studios ($38k for the 4-unit alt). Custom license mandates attribution and a separate deal past $20M revenue. Scores 39 on the Artificial Analysis index, behind GLM 5.3 (45) and Kimi K3 (44).
- NVIDIA Nemotron 3.5 Bolt: 74B total/9B active, quantization-aware training down to 2bit — 22GB + 1.7GB vision projector, deliberately built for 24GB cards, with day-one Unsloth GGUF, llama.cpp, Ollama and LM Studio support.
- Reflection Beam: 501B total/23B active, Apache 2.0, smallest checkpoint NVFP4 at 294GB.
- DeepSeek V4.1 Flash 1017: MIT license, vendor-reported DeepSWE 77.0 vs Opus 5's 74.0, 510GB repo.
Author's takeaways: 'runs locally' now means a computer that costs more than a car; licenses mattered more than parameter counts this week; Mistral Large 4 (1T) lands by month's end.
More from Infra
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Jensen Huang lays out NVIDIA's full Vera Rubin agentic system: NVL72, MGX, CMX and RTX — rohanpaul_ai · 2026-10-23
- ONNX to Transformers.js: running models directly in your browser — JosephJacks_ · 2026-10-23
- MiniMax M3.1 Flash free for 3 days on 0G, with 1M context and 512K output — MiniMax_AI · 2026-10-23
- "We bet on Cloudflare and it paid off": R2 and D1 at billion scale — dinasaur_404 · 2026-10-23
- Public data is drying up; private training data is turning into a browsable catalog — SucceededMind · 2026-10-23