Databricks rolls out GPT-6 Astra to 3,500 engineers, coding spend jumps 60%
nickbaumann_ · x · 2026-09-17
Databricks engineering lead Patrick Wendell shared internal results after rolling out GPT-6 Astra to all 3,500 engineers:
- On highly complex tasks—especially high-level system design and long-range horizontal work—Astra unambiguously beats their previous top models (Claude Opus 5, Sol 5.6).
- A pilot with 200 users showed overall coding spend rose about 60% vs baseline. Powerful, but expensive.
- No meaningful gains on medium/low complexity tasks; the team suspects those are already saturated by existing models.
More from Companies & People
- Ex-HRT quant: LLM agents now let average CS grads do elite quant research — igarciacamargo · 2026-09-17
- OpenAI Reportedly Pours Resources into 3D/4D as Large-Scale Real2Sim Nears — akanazawa · 2026-09-17
- Harvey builds founding research team, bets on open-weight models for legal AI — sunjiao123sun_ · 2026-09-17
- OpenAI warns industry hasn't solved alignment enough to keep scaling at max speed — Polymarket · 2026-09-17
- Supabase Select 26 lineup: YC's Garry Tan and Anthropic execs to speak in SF — garrytan · 2026-09-17
- Databricks rolled Astra out to all 3,500 engineers: beats Opus 5 on complex tasks, +60% coding spend — gdb · 2026-09-17