dots3-note Preview: 280B Open-Source Multimodal Model for Agency
iamrobotbear · x · 2026-08-17
dots3-note preview has been released, an open-weight multimodal model designed for long-horizon agency in real life.
- Specs: 280B total parameters, 16B active parameters.
- Context: Supports 512K context window.
- Modalities: Text + Vision + Audio.
- Capabilities: Built to reason, use tools, learn, and adapt over time.
Related event: dots3-note Preview: 280B Open Multimodal Model for Long-Horizon Agents(3 posts)→
More from Multimodal
- Building characters straight from MiniMax H3 T2V + R2VA, no more SDXL anchors — SIR_NVAX_A_LOT · 2026-08-17
- An entire music video made with MiniMax H3 visuals, Suno music, and DaVinci Resolve — la_art · 2026-08-17
- LTX-2.5 vs MiniMax H3 i2v on an RTX 5090: 1080p vs 1344×768 is what 32GB fits — chanteuse_blondinett · 2026-08-17
- Demo: Seedance 2.0 generates seamless long-form narratives in PixVerse — umesh_ai · 2026-08-17
- Mureka launches V9.5 AI music model, claims to solve "sounds fake" problem — Aiden_Tech_Ai · 2026-08-17
- Matrix meme remake: "All I see is Seedance, Omni, FLUX, H3" — DavidmComfort · 2026-08-17