Leaked ByteDance Doubao Dots3-note Specs: 280B Total Params, 512K Context
teortaxesTex · x · 2026-08-14
A blogger leaked the suspected specs for ByteDance's new Doubao model (Dots3-note) on Xiaohongshu. According to the image, the model has 280B total parameters with 16B active, supports a massive 512K context length, utilizes 13 DSA + 33 SWA attention mechanisms, and accepts text, image/video, and audio inputs.
In terms of evaluations, while it loses to existing models on "mundane" evals, it claims impressive scores on ARC-AGI tests: scoring 81 on ARC-AGI 2 and 32.1 on ARC-AGI 3. If real, it could be a highly competitive model.
Related event: Xiaohongshu Open-Sources dots3-note 280B Multimodal Model(14 posts)→
More from Models
- From botching 9.9 vs 9.11 to tackling the hardest math problems in two years — Yuchenj_UW · 2026-10-07
- OpenAI claims 372 unsolved problems cracked, averaging about 3 hours each — i_dg23 · 2026-10-07
- Benchmark: OpenAI Decisions API costs 2x more, 5-10% worse than Jev — xeophon · 2026-10-07
- Cagliostro V3.5 135M Dethrones SmolLM2-135M on Open SLM Leaderboard at 27.49 — Megneous · 2026-10-07
- "If you still think LLMs suck at math, it's a skill issue" — hot take gains traction — alejandroll10 · 2026-10-07
- Sauers_: Opus 5.5 shows fewer strong technical opinions than Fable 5.1 and 5 — Sauers_ · 2026-10-07