Add benchmark results for openai/gpt-5.5-pro

[github-actions]

This PR adds benchmark results for the openai/gpt-5.5-pro model.

Results have been pushed to Tinybird with validated=0 (pending review).
Merging this PR will validate the results, making them visible on the production dashboard.

This PR was automatically generated by the benchmark workflow.

Note: If you don't want to merge this PR, close it and the model will be added to the failed list.

/cc pei-tinybird,gnzjgo

[github-actions]

Automated Benchmark Review

Model: openai/gpt-5.5-pro
Review model: openai/gpt-5.4-nano
Success rate: 20.0%

---

1. Quality summary: The run shows low overall reliability: only 10/50 requests succeeded (20% success rate). While latency and execution time are not extreme (avg latency 173 ms, avg execution time 2.16 s), the high error rate suggests frequent failures rather than performance degradation.

2. Concerns/anomalies: The primary anomaly is the very high error count (40/50) with firstAttemptRate = successRate = 20%, indicating no meaningful recovery on retries. This points to consistent issues (e.g., prompt/format mismatch, integration failures, rate limits, or model/tooling errors) that warrant investigation before relying on these results.

3. Recommendation: REVIEW.
   Recommendation: REVIEW

---

This review was automatically generated. Set AUTO_MERGE=false in repo variables to disable auto-merge.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
llm-benchmark	Ready	Preview, Comment	Jun 18, 2026 10:04am

[github-actions]

Model Rejected

The model openai/gpt-5.5-pro has been added to the failed models list.

What this means:

• This model is tracked as a rejected benchmark
• Discovery will still find it, but it won't be re-benchmarked automatically while in the failed list

To manage this model:

# Remove from failed list (to allow re-benchmarking)
npm run manage-failed-models remove openai gpt-5.5-pro

# List all failed models
npm run manage-failed-models list

@alrocar