WHAT NEW IN X11.03 TUE

AI collections from X · 22 collections · 66,263 posts

gpt-6IMAGE

Setting GPT-6 Astra to a 1M context window in Codex

The author recommends manually setting Codex's context window to 1 million tokens and enabling the context usage display, saying Astra still has 92% recall above 500K tokens.

LinearUncle
@LinearUncleArchived 2026.09.17Originally posted 2026.09.16
Setting GPT-6 Astra to a 1M context window in Codex (image 1 of 2)
Setting GPT-6 Astra to a 1M context window in Codex (image 2 of 2)
LinearUncle @LinearUncleOriginally posted 2026.09.16
Open original on X

gpt-6-astra上下文在codex 里频繁被自动compact压缩真的影响编程效果! 我还是坚持打开模型的百万上下文配置: model = "gpt-6-astra" model_context_window = 1000000 另外建议在设置->general->composer里把“show context window usage”开关打开,这样可以看到当前的上下文占用量。 AI编程不要人云亦云,要有自己的判断,好的coding力工模型例如deepseek-flash, glm-5.3-flash, opus都是百万上下文,fable更不用说了,anthropic的东西虽然贵,但是成功的独立开发者都在使用它。 openAI默认那么点上下文更多是从成本上考虑的,毕竟它的客户成分更杂,少一些成本省很多。 我查了下实测,astra模型在超过500K以上召回成功率仍然高达92%,是很适合百万上下文的。

x.com/linearuncle/status/2100147093860151794Archived 2026.09.17 10:19