Antigravity MODEL_CAPACITY_EXHAUSTED vs Quota

Fix an Antigravity capacity error by reading the entire message, then open /usage (alias /quota). A reset timer or exhausted baseline points to quota handling; “no capacity,” “high demand,” or MODEL_CAPACITY_EXHAUSTED points to temporary model availability. Do not diagnose from HTTP 429 or 503 alone.

Diagnose the Full Message

Antigravity users have posted errors such as the following in the Google AI Developers Forum:

RESOURCE_EXHAUSTED (code 429): You have exhausted your capacity on this model. Your quota will reset after ...

UNAVAILABLE (code 503): No capacity available for model ...

Those strings are user-reported examples, not a complete official error contract. Google's official CLI documentation gives you the authoritative diagnostic tool: /usage refreshes model configuration and quota status from the backend and displays remaining quota per supported model.

/usage
# alias
/quota

Fix a Quota or Reset-Timer Error

1. Open /usage and identify the affected model.
2. If the error gives a refresh time, stop retrying until that time. Repeated retries do not replenish quota.
3. On Google AI Pro or Ultra, review the AI Credit Overages setting. The official Plans page says purchased or promotional credits can cover usage above baseline quota when the setting is enabled.
4. If another model has available quota and fits the task, select it with /model.

Google says Pro and Ultra baseline quota refreshes every five hours until the weekly limit is reached; accounts without those plans receive a weekly refresh. It also says usage is correlated with the amount of work the agent performs, so prompt count alone is not a reliable quota measure. See the current Antigravity Plans documentation.

For a deeper explanation of weekly limits, continue with the weekly quota and cooldown guide. For workload reduction, use the quota-saving workflow. This page stays focused on identifying the error branch.

Fix a Capacity or High-Demand Error

1. Preserve your prompt or work state before retrying.
2. Wait briefly and retry once. A capacity shortage is not repaired by changing local credentials.
3. Open /model and choose another model that your plan offers. The official model picker is sticky within a conversation, and an in-flight turn keeps its previous model until it finishes or is cancelled.
4. If every offered model fails, pause and check support channels rather than looping requests.

Google's Gemini API troubleshooting page defines 503 UNAVAILABLE as a service that may be temporarily overloaded or down. That is a general Gemini backend definition, not a guarantee about every Antigravity error, but it supports retrying availability failures later rather than deleting local state. See the official Gemini API troubleshooting guide.

Why a 429 Can Be Ambiguous

Google's Gemini API guide associates 429 RESOURCE_EXHAUSTED with rate limits. Separately, Google Cloud's Vertex AI documentation says a 429 can also mean that requests exceeded available processing capacity in a shared resource pool. Antigravity's own Plans page says baseline limits are partly determined by capacity. The safe conclusion is to use the full error payload and Antigravity's live quota panel, not the number alone.

• Official Gemini API error definitions: 429 and 503 troubleshooting
• Official Vertex AI behavior: error code 429

Avoid Ineffective or Risky Fixes

• Do not delete Antigravity configuration. Quota and backend capacity are not restored by removing local settings.
• Do not repeatedly log out and in. Reauthentication is appropriate for an authentication error, not a message that names quota or model capacity.
• Do not route your Antigravity login through a proxy. Google's FAQ says third-party access to Antigravity violates its Terms and may lead to account action.
• Do not assume BYOK will bypass quota. The Plans page says Antigravity does not support bring-your-own-key or bring-your-own-endpoint for additional rate limits. See the BYOK support guide.

Escalate with Useful Evidence

If the same error persists after the displayed reset time or across all offered models, capture:

• the complete error text and timestamp;
• the model selected;
• a screenshot of /usage with private data removed;
• whether AI Credit Overages is enabled;
• whether the same failure occurs in a new conversation.

Use Google's Antigravity support page for current support channels. Do not post account tokens, authorization headers, or unredacted logs.

Frequently Asked Questions

Does 429 always mean I used all my quota?

No. Google documents 429 for both rate-limit and shared capacity cases across its generative AI services. Read the message details and open /usage.

Should I switch models?

It is a supported diagnostic and workaround when another offered model has capacity or remaining quota. Use /model; do not assume every model has an independent quota.

Can AI credits continue a blocked session?

For eligible Pro and Ultra users, credits can cover overage after baseline quota when enabled. Credits do not guarantee capacity during a model-availability incident.

Will reinstalling Antigravity fix it?

Not when the error explicitly names quota or model capacity. Reinstall only for a separately diagnosed installation problem.

All Sources and Links

Primary Google documentation

• Antigravity CLI Model Quotas (/usage and /quota)
• Antigravity Plans — refreshes, weekly limits, and overages
• Antigravity Models — current model availability
• Antigravity CLI reference — /model and quota commands
• Gemini API troubleshooting — 429 and 503 definitions
• Vertex AI error code 429 — shared-capacity behavior

Community report used only for the observed strings

• Google AI Developers Forum — user-reported Antigravity 429 and 503 messages