Blogger: Chinese labs have cracked scaling and RL, need ~6 months to reach Astra level
zephyr_z9 · x · 2026-09-06
Commenting on the gap between Astra and Chinese models, the poster argues Chinese labs have already solved parameter scaling and RL, and will go all-in on model depth next. He estimates they need at least 6 months to reach Astra-level, by which time OpenAI will ship "Bel". Unverified personal take.
More from Models
- OpenAI insider: 'Astra genuinely shocked me' and 'ARR doesn't matter anymore' — Yuchenj_UW · 2026-09-06
- AI editing a math paper spots a counterexample to a proposition the author planned to cite — jessi_cata · 2026-09-06
- Dev calls GPT-6 Astra influencer-first rollout annoying but says it's the best model yet — AIandDesign · 2026-09-06
- Is GPT-6 Astra painting a portrait in Canva actually unimpressive? A Reddit breakdown — meh_coder · 2026-09-06
- New Codex compaction looks like a message board — is that why models love leaving notes? — mariofilhoml · 2026-09-06
- Kimi K3 beats GPT-6 on game menu replication quality but takes 5x longer, $3.75 vs $2.30 — ChrisGPT · 2026-09-06