MOSS AMA highlights: image models, local speech, audio/video research and open-source culture
Physical_Hat4022 · reddit · 2026-11-04
A Reddit user shares highlights from MOSS team's October 30, 2026 AMA covering image models, local speech, audio and video research, and open-source culture. Screenshots show the team attributing their quality to sweating every detail from data to training and inference, and member Gary crediting Shanghai's dev culture of open sharing and peer recognition for the local hacker community energy.
More from Models
- Weight-editing author: gains aren't real knowledge — Gemma math didn't improve, but Qwen3 issue is real — lateinteraction · 2026-11-04
- rasbt questions Qwen3 weight-editing gains: 50 GRPO steps already lift MATH-500 from ~15% to ~50% — rasbt · 2026-11-04
- NVIDIA releases Dory1.0-4B, a looped recurrent LM beating larger models — nvidia · 2026-11-04
- New GDPx Benchmark: Frontier Models Score Near 0% on Full Legal Jobs — jkubicki · 2026-11-04
- Claude Shows Self-Awareness About Its Copyright Lyric Refusals — Angaisb_ · 2026-11-04
- Claude Opus 5 surgically edits ~1% of ProteinMPNN weights into SolubleMPNN-like behavior — AllThingsApx · 2026-11-04