Rednote open-sources dots3-note: 16B active params hits 78.4% on SWE-bench
heyshrutimishra · x · 2026-08-31
Rednote released open-weight dots3-note preview: 280B total / 16B active parameters, 512K context, unified text-vision-audio, Apache 2.0.
Benchmarks: ARC-AGI-2 81.39, SWE-bench Verified 78.4%, TerminalBench 2.1 75.1%, MMMU Pro 79.1%.
The story is training, not architecture: a new RL method called TEMPO fixes credit assignment in tasks that run for hours rather than seconds. Dropped into unseen environments, the model explores, forms hypotheses, verifies them, and rewrites its own memory — memory isn't hand-designed; RL learns what's worth remembering. The team also open-sourced VibeSearchBench and VibeLifeBench, where every frontier model currently fails: long-horizon agency remains unsolved.
More from Models
- NVIDIA's Nemotron 3.5 Super VL hits CoreWeave: 12B-active multimodal MoE at $0.25/M input tokens — wandb · 2026-10-16
- Opus 5 might be a very deep model or use looping, commentator speculates — scaling01 · 2026-10-16
- GPT-6.1 Sol audit finds 6% of arXiv papers had false main theorems, down to 1.5% recently — nitarshan · 2026-10-16
- Nvidia-backed post-training pushes Nemotron 3.5 VL past frontier on dermatology diagnosis — LoganGrasby · 2026-10-16
- Gemini 4 Argon scores 97.5% on ARC-AGI-3, second lab above 95% — patience_cave · 2026-10-16
- Microsoft MAI claims SOTA 3T-param in-house base models, post-training results coming — soldni · 2026-10-16