Thinking Machines ships v2 small model: 68-hour pretraining on GB300, NVFP4 throughout
SonglinYang4 · x · 2026-11-06
Soumith Chintala announced Thinking Machines' v2 small model: pretrained in just 68 hours on GB300s, with post-training via 17 million rollouts over 9 days, fully pre-mid-post trained in NVFP4. It's available to train and fine-tune on Tinker now, with open weights coming next week after external safety testing.
More from Models
- Claude hits conversation length limit after ~10 messages, user flags Anthropic — ___Patrice___ · 2026-11-06
- Researcher speculates on mystery model: LatentMoE, linear attention, 12T tokens — eliebakouch · 2026-11-06
- Yacine publicly asks insiders how many GPUs pretrained Inkling Small 2 — yacineMTB · 2026-11-06
- Autonomous AI R&D Runs 'Extremely Dumb Experiments', Researcher Reports — panickssery · 2026-11-06
- Claude Opus 5 pulls off multi-level roleplay: staying in character while knowing it's acting — repligate · 2026-11-06
- Inkling 2 Small hits 46.8% Pass@1 on APEX-Agents, up 13 points over predecessor — ziqiao_ma · 2026-11-06