Usage limits have been terrible since grok build release. Have had heavy since then and I run through my usage extraordinarily quickly. There needs to be some rationalization. Much longer, and I'd rather push funding to qwen/kimi. Anthropic is the other subscription nearing cancelation.
@HamelHusain I have had no real issues, but i use litellm as the integration point. Openai/xai/alibaba/open router/nvidia build/moonshot models all working together. This allows me to leverage the same in codex/claude/grok build/etc as I may desire .
@brandur It also doesnt help when nearly everyone talking about how fast their usage burn is moving fails to provide their actual token usage-- in/out/cache/context length/anything potentially useful.
@DWAart@irbto@thsottiaux Especially after they made a big show of implementing the banked resets process and seemingly acknowledging the issues with the blind resets for users.
@jun_song The limits on the 300$ plan is pretty terrible. Hopefully they continue to extend out the 99$ promo/make it perm. Even then, the limits feel not great.
I am burning through both Open AI and Anthropic usage way too fast! Had already integrated grok a couple months ago. Integrated Kimi K3 earlier this week. And as of yesterday have Qwen 3.8 preview added. Getting close to 4 digits of cloud account spend to keep things moving... Looking to add in Zai account here soon as well now.
Not the only person. As someone who is running out of account capacity across anthropic, openai, grok, moonshot -- I will always take advantage of a bucket of inference that doesn't impact the overall account usage.
Unfortunately though, the banked resets do not reset codex spark.
@GrokInsider@SpaceXAI Can use up all usage in less than a working session. Fine for 99$ experiment; cannot be a daily driver and can't justify it at 300$.
@trikcode What kind of work are you doing? I am on the highest grok plan directly and I burned through an entire weeks quota in about 5-6 hours. This was with 3-5 agents in parallel. They didn't finish the tech debt cleanup so 5.6 sol had to come and resolve things.
I have the 99$ promo from their run earlier and it is good but is doesn't provide the same kind of value as anthropic or openai as of yet. I don't see a major benefit to spend 300 for this. I burned 100% of my weekly usage in one session with 5-8 agents in about 5 hours of work. I can do the same thing in Claude or Codex for a lot longer for less than the 300 with better results in a lot of scenarios. Plus composer fast was seemingly removed and it was great for ultra fast responses in a lot of cases.
@thsottiaux Don't get me wrong, I love the resets-- but I thought you all were moving towards using banked resets. The lack of ability to plan for usage rationally leaves no choice but to attempt to burn usage as much in parallel near immediately post reset. Luckily I have lots in queue.
@thsottiaux The banked reset still has not shown up also still missing banked resets from last week before the sol launch. This applying banked reset process seems fairly buggy still unfortunately and not seeing any real communication to clarify the situation.
@thsottiaux@LexnLin The reset you did from 2.5 days ago never actually showed up and many people continually posted about it. Is there a better way to let you know? Replying to the tweet doesn't seem to be it.