GPT-6 Astra criticized as poor performer; Kelsey Piper calls it smartest but least useful model
Tenobrus says GPT-6 Astra is "genuinely a very bad model" that ignores user intent, comparing it to the rocky Opus 5 rollout. Kelsey Piper pushes back: Astra is the smartest model but not the most useful for underspecified tasks; Tenobrus agrees it's very smart. No safety-specific claims raised.
All posts · 5, most important first
- upon further experience with it, and in comparison with opus 5.5, astra is genuinely a very bad model. i find myself getting dramatically more done with normal-speed opus vs *ultrafast* astra. i find it tough to even use as a subagent for opus. it just really really does not care about user intent.9 likes · 0 reposts
- I think Astra's the smartest model but it's definitely not the most useful model for underspecified tasks.9 likes · 0 reposts
- i agree that it's very very very smart5 likes · 0 reposts
- this increasingly feels like an opus 5 situation man. they gotta fix this shit asap.3 likes · 0 reposts
- sol 6.1 is pretty pareto optimal on cost and speed for small well scoped tasks, so that's nice, gives me some way to use openai tokens. but astra 6.1 better land like yesterday4 likes · 0 reposts
@tenobrus commentary · 
@kelseytuoc press ·