Rerun GPT 5.6 Luna/Terra at new pricing

Add topic breakdown for Qwen2.5-Coder-32B

Add Qwen 2.5 Coder 32B, Qwen 3 4B Instruct 2507, Deepseek R1 0528, and Devstral Small 2507

Add Qwen 3 Coder 30B-A3B, Claude Sonnet 4, Gemini 2.5 Pro, Gemini 3.1 Pro, GPT 5

Breakdown QA scores by topic

Add fail/noCode to leetcode results and the exact release date of the model

Rerun benchmarks with more robust retries for fewer no_answer failures

Adjust price according to OpenRouter effective pricing

Better breakdowns and table layout

Add Kimi K3

More precise data entry with less rounding

Add Claude Opus 5

Add Grok 4.5 (Medium)

Add GPT 5.6

Add Qwen 3.5 9B

Use the same rounding as everyone else

Improve tie behavior

Add Claude Fable 5

Add llama 3.1 8b

Initial commit