Amp adds GPT-6 Astra with ChatGPT subscription usage
Coding tool Amp now offers GPT-6 Astra, where its high mode maps to Astra's medium effort level, and users can customize reasoning effort.
Amp now has GPT-6 Astra. It's very good and uses your linked ChatGPT sub. Amp's `high` mode is gpt-6-astra medium (not xhigh, why in thread). In other Amp modes soon. BTW: customize your Amp dial at https://ampcode.com/settings/dial or pick any GPT-6 Astra effort level in Raw Models
Re: why our high mode uses gpt-6-astra medium: We've found GPT-5/6 models' high/xhigh effort levels to overthink, taking way longer and using more tokens than you'd want. Instead, Amp gives the agent the oracle, which lets it go deep when truly needed, but otherwise not overthink. The oracle is a subagent (http://ampcode.com/modes) with an intentional and adversarial system prompt and/or higher effort level that. We were actually quite close to temporarily deprecating Amp's `high` mode until we got access to GPT-6 Astra, for this reason. High mode was so much slower without meaningfully better results. Also, why no GPT-6 Astra in Amp's `medium` mode yet? It will get there soon, like before the end of the weekend. We want to let rate limits and such settle over the next day or so and test a few other things before making the switch. (And remember, you can now customize your dial's `medium` mode in the link in the parent post.)