K2 Horizon researcher: true open source means data, code and recipes, not just weights
HongyiWang10 · x · 2026-09-06
K2 Horizon researcher Hongyi Wang used the release to argue that true openness goes far beyond downloadable weights: it should include training data, code, recipes and enough detail for others to reproduce, adapt and improve the model. K2 Horizon is a connected family of six foundation models spanning 0.9B to 375B parameters, delivering top-tier performance in coding and agentic tasks at every size class.
Related event: K2 Horizon Open-Sources Six Models from 0.9B to 375B(2 posts)→
More from Models
- Kimi K3 beats GPT-6 on game menu replication quality but takes 5x longer, $3.75 vs $2.30 — ChrisGPT · 2026-09-06
- User tests OpenAI Astra computer use: one PPT slide took nearly 15 minutes — bookwormengr · 2026-09-06
- Claude Navier-Stokes proof rumor officially debunked by Elliot Glazer — basedjensen · 2026-09-06
- OpenAI's GPT-6 Astra touted as its most advanced model: autonomous browsing, building, tasks — CreamHoliday4754 · 2026-09-06
- Dev: Astra nails in one shot what took days of fighting with Opus — gabrielchua · 2026-09-06
- Users report GPT-6 Astra finding up to 176x code speedups — or nothing at all — ivan_bezdomny · 2026-09-06