v0.8.0 — No Retries

[0.8.0] no retries

Changed

• Error Handling Strategy: Completely removed retry mechanisms in favor of fail-fast approach
• Removed exponential backoff and retry logic from OpenRouter client
• Simplified API error handling to immediately move to next model on failure
• Changed default retry count from 1 to 0 for all benchmark operations
• Updated run_benchmark_on_model() and benchmark_single_model() to not retry on empty responses
• Sequential Execution: Enhanced progress tracking and error handling in sequential mode
• Improved progress bar accuracy by adjusting total count for failed models
• Better error reporting with clear model-by-model failure messages
• Streamlined sequential execution flow without retry overhead

Updated

• Leaderboard Results: Latest benchmark rankings with new top performer
• New leader: meta-llama/llama-3.3-70b-instruct:free (1640.68 points, 36.3s)
• Added new failed model: cognitivecomputations/dolphin-mistral-24b-venice-edition:free
• Updated rankings reflect improved performance testing methodology

Improved

• Performance: Eliminated retry delays for faster batch processing
• Reliability: Clear failure detection without hanging on problematic models
• User Experience: More predictable execution times and clearer error feedback