I’m experiencing a persistent “high traffic” error that has made my Antigravity account unusable.
I had never encountered this issue before. It started immediately after I ran a task that lasted about 5 minutes. Since then, the “high traffic” message keeps appearing repeatedly, and I can no longer use the product normally.
Other people in the same family group do not have this problem and can use Antigravity normally. So this does not look like a real global high-traffic issue. It seems more like an account-level restriction or throttling being shown as “high traffic.”
Please check whether my account has been rate-limited, flagged, or restricted, and let me know how to restore normal access.
Hi @Bigboom,
To help us better assist you and investigate this, could you please provide a few additional details?
- The version of Antigravity you are using (e.g., 2.0, IDE, or CLI, and the specific version number)
- Your operating system
- The type of task you are running (for example, does it involve processing a high volume of workspace files, running an intensive multi-file refactor, or executing a tight loop within the terminal?)
- Whether this block triggers across every available model, or only specific ones
- Any sample prompts you are using for your task
Hi Siddharth,
Here are the additional details:
- Antigravity version: IDE mode, version 1.23.2
- Operating system: Windows 11 25H2, build 26200.8246
- Task type: I am using Antigravity to modify, supplement, and reorganize a large Markdown research note file. The note is around 1,400 lines / 100 KB.
- Context: The note records my reading process for a research paper. Since the content was added multiple times over time, its structure is somewhat disorganized. I asked the agent to read the Markdown note, reorganize the content based on the existing material, improve the flow, and make the writing more coherent. I also asked it to supplement missing proof details from the original paper.
- Model affected: This happens specifically with Opus. I chose Opus because it performs better for research/math-related reasoning tasks.
- Sample prompts:
The original prompt that first triggered the high traffic block was:
Chinese original:
“你把原文定理2及定理3的证明过程给详细补充进去。”
English translation:
“Please supplement the proof processes of Theorem 2 and Theorem 3 from the original paper in detail.”
Later, the following prompts repeatedly triggered the high traffic block:
- Chinese original:
“笔记内容太乱，你重新整理。”
English translation:
“The notes are too messy. Please reorganize them.”
- Chinese original:
“你先列个执行计划。”
English translation:
“First, please list an execution plan.”
- Chinese original:
“另外，我觉得原第5节之后的内容，该合并为同一章节就合并为同一章节，不要搞那么多章节。”
English translation:
“Also, I think that after the original Section 5, any content that should be merged into the same chapter should be merged accordingly, instead of creating too many separate sections.”
- Chinese original:
“问题1，改为3.2。问题2，更新。”
English translation:
“Change Question 1 to Section 3.2. Update Question 2.”
At this point, almost any input seems to trigger the high traffic block. For example, even the warm-up step at the beginning of each conversation, where the agent reads my memory file, can trigger it. Very short prompts such as “连接测试” (“connection test”) can also trigger the same high traffic block.
Currently, the forum only allows me to upload one image. The remaining images can be found at this link:
https://drive.google.com/file/d/1WHxpjAyQCicD-rTk5h8_m4n0l-UXS_o0/view?usp=sharing
Just now, I only sent a simple message saying “connection test,” yet my account did not receive a response until after more than 10 “high traffic” notices. At the same time, another account in my family group did not encounter this issue at all and received a smooth response immediately after asking a question.
Therefore, I have very good reason to believe that this “high traffic” notice is not genuine. I pay $200 per month for the subscription, yet Google is restricting my access to the model even though I am using it completely normally. I am truly very disappointed.
Another account in my family group used Claude for only about three 5-hour sessions, and now the Claude cooldown time for the entire family group has become 50 hours.
Many people on the forum have already been complaining that Google’s token counting is extremely unclear. Users have no idea what exactly happened or how the usage limit was triggered, and the software itself is also full of bugs.
What makes it even worse is that Google still does not provide real human customer support to handle customer complaints. Instead, it keeps asking us to post messages on the forum.
I am truly, deeply disappointed with Google. This is by far the worst Google product I have ever used, without a doubt. I have already canceled my Google Ultra 20x subscription, and I strongly advise anyone who sees this message and is considering subscribing to a Google AI plan to completely abandon that idea, because it will definitely make you feel extremely regretful and make you feel that it is absolutely not worth the $200 you paid.
