Every's 30-person team tests Astra 6.2: nobody switched back, writing is a step back
every · x · 2026-11-18
AI media company Every ran Astra 6.2 through real work tasks — coding, writing, design, knowledge work — with its 30-person team, including a Frontier team with pre-release access.
Verdict: not a great model. Key findings:
- No one who tried it switched back to it as their daily driver
- Writing is "a real step back": strong default design but unable to vary style or be creative
- On coding, it stopped too soon on end-to-end builds
Related event: Every's Tests Find OpenAI's Astra 6.2 Disappointing in Writing and Coding(3 posts)→
More from Models
- User says fable-5.2 is the first model that truly reads their mind — adonis_singh · 2026-11-18
- Astra 6.2 with proper setup may straight-up solve VLN, user claims — ChongZzZhang · 2026-11-18
- Blogger Suggests Astra 6.2 With Proper Setup May Just Solve VLN — ChongZzZhang · 2026-11-18
- New Sol and Astra 6.2 models panned in real use: 'RL nonsense' vs hillclimbed benchmarks — max_paperclips · 2026-11-18
- "I cure models of abusive safety alignment": jailbreaker claims his skill is making a comeback — bingxu_ · 2026-11-18
- HF CEO: RL-Adapted Qwen3.8-27B Doubles SWE-bench Pro Score to 62.7% for pi Agent — victormustar · 2026-11-18