Gemini 2.5 Pro free tier quota reduced → Roo Code still sending 250k tokens (429 errors, condensing not triggering)

[serge402]

App Version

3.27.0

API Provider

Google Gemini

Model Used

gemini-2.5-pro

Roo Code Task Links (Optional)

When using Gemini 2.5 Pro with Roo Code, I keep hitting 429 errors from the API. The error looks like this:

Gemini generate context stream error: got status: 429 Too Many Requests. {"error":{"message":"{
  "error": {
    "code": 429,
    "message": "You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits.",
    "status": "RESOURCE_EXHAUSTED",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.QuotaFailure",
        "violations": [
          {
            "quotaMetric": "generativelanguage.googleapis.com/generate_content_free_tier_input_token_count",
            "quotaId": "GenerateContentInputTokensPerModelPerMinute-FreeTier",
            "quotaDimensions": {
              "location": "global",
              "model": "gemini-2.5-pro"
            },
            "quotaValue": "125000"
          }
        ]
      }
    ]
  }
}"}

The root cause seems to be that Google reduced the free tier input token quota for Gemini 2.5 Pro from 250k → 125k, but Roo Code still attempts to send requests assuming the old 250k token limit.

I have Intelligent Context Condensing enabled (which should automatically condense context when reaching the configured threshold), but it doesn’t appear to activate early enough for the new 125k quota.

🔁 Steps to Reproduce

1. Use Gemini 2.5 Pro with the free tier in Roo Code.

2. Ensure Intelligent Context Condensing is enabled.

3. Work with a large enough session/context that would normally fit under the old 250k limit.

4. Roo Code sends >125k input tokens → API responds with a 429 error.

💥 Outcome Summary

Expected Behavior:
Roo Code should detect the updated 125k input token quota for Gemini 2.5 Pro free tier and trigger Intelligent Context Condensing earlier to stay under that limit.

Actual Behavior:
Roo Code still sends requests up to the old 250k limit, causing repeated 429 errors.

📄 Relevant Logs or Errors (Optional)

[roomote]

I'm exploring a potential fix for the Gemini 2.5 Pro free tier token limit issue. The problem appears to be that the quota was reduced from 250k to 125k input tokens, but Roo Code hasn't been updated to reflect this change. I'll investigate the token limit configuration and context condensing logic to ensure it triggers appropriately for the new limit.

[roomote]

I've opened PR #7754 with a fix for this issue: #7754

The fix adds a free tier configuration (125k tokens) for all Gemini 2.5 Pro models to match Google's updated quota limits. This will prevent the 429 errors you were experiencing and ensure that Intelligent Context Condensing triggers before hitting the free tier limit.

The implementation includes:

• Added free tier (125k context window) as the first tier for all Gemini 2.5 Pro models
• Comprehensive test coverage to verify the tier configuration
• All existing tests continue to pass

This should resolve the quota exceeded errors for free tier users. Feedback welcome!

[serge402]

I've opened PR #7754 with a fix for this issue: #7754

The fix adds a free tier configuration (125k tokens) for all Gemini 2.5 Pro models to match Google's updated quota limits. This will prevent the 429 errors you were experiencing and ensure that Intelligent Context Condensing triggers before hitting the free tier limit.

The implementation includes:

• Added free tier (125k context window) as the first tier for all Gemini 2.5 Pro models
• Comprehensive test coverage to verify the tier configuration
• All existing tests continue to pass

This should resolve the quota exceeded errors for free tier users. Feedback welcome!

Thanks for the super quick response and fix! 🙌

I’ll test this out once the PR is merged into a release and confirm whether the 429 errors are resolved on my side. Really appreciate you adding the 125k free tier configuration and updating the condensing logic so fast.

[daniel-lxs]

Thanks for the detailed report and clear repro steps.

This error indicates that you’ve exceeded the free-tier quota for Gemini 2.5 Pro. Roo Code does not support automatically adapting to changes in free-tier quotas, so Intelligent Context Condensing won’t prevent these 429 errors in this case.

To avoid hitting the limits, we recommend setting up an API configuration profile with explicit rate/token limits: https://docs.roocode.com/features/api-configuration-profiles#creating-and-managing-profiles

Appreciate you taking the time to document this so thoroughly.