# Error giemini api key

**URL:** <https://discuss.ai.google.dev/t/error-giemini-api-key/202677>\
**Category:** Gemini API\
**Tags:** api-key\
**Created:** [November 19, 2026, 2:27pm UTC](https://discuss.ai.google.dev/t/error-giemini-api-key/202677 "2026-11-19T14:27:03Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![ABDELAZIZ\_BENYAHIA](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/abdelaziz_benyahia/32/160380_2.png) [@ABDELAZIZ\_BENYAHIA](https://discuss.ai.google.dev/u/ABDELAZIZ_BENYAHIA)\
**Post date:** [November 19, 2026, 2:27pm UTC](https://discuss.ai.google.dev/t/error-giemini-api-key/202677/1 "2026-11-19T14:27:03Z")

</div>

Subject: Gemini API Free Tier — gemini-3.6-flash: repeated 503 + extended “cooling down” + hard timeouts (reproducible over 2 days) + need guidance on restricting AI-Studio-issued key

Hello,

I’d like to report a reproducible pattern of availability issues with gemini-3.6-flash on the Gemini API Free Tier. Full details below.

ACCOUNT / PROJECT INFO

- Google Cloud / AI Studio project name: [\_\_\_\_\_\_\_\_\_\_\_]

- Project ID: [\_\_\_\_\_\_\_\_\_\_\_]

- API key: created via aistudio.google. com/apikey, last 4 characters: [\_\_\_\_]

- Billing account attached: NO (Free Tier only)

ENVIRONMENT

- Model: gemini-3.6-flash

- Client: calls routed through OmniRoute (open-source local API gateway), v3.8.52, on http:/127.0.0.1:20128

- Device: Android phone, via Termux — no desktop/server

- Request pattern: manual, single-prompt test calls only, ~5-10 total over 2 days

EXACT REQUEST SHAPE

{

“model”: “gemini/gemini-3.6-flash”,

“messages”: [{“role”: “user”, “content”: “”}],

“max\_tokens”: 100-400

}

ERRORS OBSERVED (same device, same key, same 2-day window)

1. Hard timeout, zero bytes returned after 50s:

2. 503 with retry hint:

3. Extended cooldown, same session:

WHAT I’VE RULED OUT

- Not network/connectivity: other providers (Groq, Cloudflare Workers AI) called from the same device/session succeed in under 2 seconds.

- Not a malformed request: same shape works fine for other providers.

- Not a local proxy bug: errors are passed through verbatim from Gemini’s backend.

I also saw the banner on this forum about restricting unrestricted API keys (effective June 19, 2026). I want to confirm my key is properly restricted, but I’m having trouble locating the restriction setting:

- My key was generated directly in Google AI Studio ( aistudio.google. com/apikey ), NOT through the standard Google Cloud Console “APIs & Services \> Credentials” page.

- When I look at that key’s details inside AI Studio, I only see: the key value, name, project name, and project number — there is no “Restrictions” or “API restrictions” field shown there.

- When I go to Google Cloud Console \> Credentials for the same project, I do NOT see this Gemini key listed at all — only an unrelated key for a different Google API (Maps Tiles) tied to a separate project of mine.

REQUEST: Could you please send screenshots or a step-by-step guide showing EXACTLY where, within Google AI Studio itself (not the general Cloud Console Credentials page, since the Gemini key doesn’t appear there for me), I can view/set the restriction status for a key that was generated directly in AI Studio? I want to be sure this isn’t contributing to the errors above.

QUESTIONS

1. Is gemini-3.6-flash currently under elevated load on the Free Tier generally, or isolated to my project?

2. Is a ~16.5-hour cooldown expected after ~10 light requests on Free Tier?

3. Is a more stable Free Tier model available right now for light/intermittent testing?

4. Could routing through a local proxy on the same device trigger stricter rate-limiting than calling the API directly?

This is early-stage personal/hobby development, not production traffic. Thank you for your time.

Gemini API and OmniRoute – Error Report

1. HTTP 401 – UNAUTHORIZED

- Authentication failure.

- Observed when executing “omniroute models --search gemini”.

- Possible causes: invalid or missing authentication credentials, or an OmniRoute local API authentication issue.

- Note: It has not been confirmed that Google Gemini itself generated this error.

2. HTTP 503 – SERVICE UNAVAILABLE

- Temporary service unavailability.

- Possible cause: temporary server overload or service availability issues on Google’s side.

3. HTTP 429 – RESOURCE\_EXHAUSTED

- Request quota or rate limit exceeded.

- Possible causes:

- Requests-per-minute limit exceeded.

- Free-tier quota exhausted.

- Daily or monthly usage limits reached.

- Account or billing-tier restrictions.

4. HTTP 403 – PERMISSION\_DENIED

- Access or permission denied.

- Possible causes:

- API key not authorized to access the Gemini API.

- API key restrictions.

- Google Cloud project or service permissions issues.

5. HTTP 400 / 500

- HTTP 400 – BAD REQUEST: Invalid request, parameters, request format, or model configuration.

- HTTP 500 – INTERNAL SERVER ERROR: Internal server-side failure.

-–

Additional Issues

A. JSON Response Parsing Error

Observed message:

“Expecting value: line 1 column 1 (char 0)”

- Possible cause: Empty response or a response that is not valid JSON.

- This may indicate a local response-processing issue and does not independently prove a Gemini server failure.

B. OmniRoute Environment Configuration Warnings

The following warnings appeared:

- “STORAGE\_ENCRYPTION\_KEY”

- “MANAGEMENT\_TOKEN”

- “DEEPSEEK\_API\_KEY”

Explanation: Values defined in the user’s home “.env” file take precedence over duplicate values in the global OmniRoute package “.env” file.

These are configuration warnings, not confirmed Gemini API errors.

-–

Summary

Five error categories were investigated:

- HTTP 401

- HTTP 503

- HTTP 429

- HTTP 403

- HTTP 400 / 500

Additional issues:

- JSON response parsing error.

- Environment configuration warnings.

Evidence note: HTTP 401 was directly observed in the latest test. The other HTTP errors were discussed during previous troubleshooting, but their original response logs are not currently available. Their causes remain possible explanations, not confirmed diagnoses.
