results.txt

Toolless (sorted by score)
---------------------------------------------------------
Model                   Score    P50 (s)     P99 (s)
---------------------------------------------------------
gemini-2.5-pro          8/20     16.32       61.77
gpt-5                   5/20     33.78      144.80
glm-4.5-air             3/20     30.33      266.54
gpt-oss-120b            2/20      8.57       63.07
Qwen3-235B (thinking)   2/20     50.75      152.70
gemma3-4b               1/20    161.28      312.12


With tools (sorted by score)
---------------------------------------------------------
Model                   Score    P50 (s)     P99 (s)
---------------------------------------------------------
gpt-5                  16/20    270.40      990.36
gemini-2.5-pro         12/20     37.22      134.73
gpt-oss-120b           11/20     12.81       33.87
glm-4.5-air             9/20     45.61      103.07
Qwen3-235B (thinking)   6/20    111.34      226.74
gemma3-4b               0/20    870.13     1900.00