Overall MCP vs Hardcoded Comparison
Averaged across all projects: MarkdownValidator, GameBuilder, EmailResponder
Framework-Level Comparison (All Models Averaged)
| MCP Mean (s) | Hardcoded Mean (s) | Diff (MCP-Hard) |
|---|---|---|
| 117.69 | 140.50 | -22.81s (-16.2%) |
Per-Model Summary
| Model | MCP Mean (s) | Hard Mean (s) | Diff (MCP-Hard) |
|---|---|---|---|
| DeepSeek-R1 | 230.39 | 298.55 | -68.16s (-22.8%) |
| DeepSeek-V3-1 | 97.30 | 98.90 | -1.60s (-1.6%) |
| GPT-4o-mini | 29.40 | 25.19 | +4.21s (+16.7%) |
| GPT-5 | 11.26 | 13.59 | -2.33s (-17.1%) |
| Gemini-2.5-flash | 236.06 | 299.20 | -63.14s (-21.1%) |
| Gemini-2.5-flash-nothinking | 152.47 | 183.54 | -31.07s (-16.9%) |
| Qwen3-235b | 66.92 | 66.22 | +0.70s (+1.1%) |
MarkdownValidator: MCP vs Hardcoded
Overall Summary (Averaged Across All Models)
| MCP Mean (s) | Hardcoded Mean (s) | Diff (MCP-Hard) |
|---|---|---|
| 111.78 | 124.33 | -12.55s (-10.1%) |
Per-Model Comparison
| Model | MCP Mean (s) | Hardcoded Mean (s) | Diff (MCP-Hard) |
|---|---|---|---|
| DeepSeek-R1 | 129.20 | 147.98 | -18.77s (-12.7%) |
| DeepSeek-V3-1 | 10.69 | 10.29 | +0.40s (+3.9%) |
| GPT-4o-mini | 3.58 | 8.47 | -4.89s (-57.8%) |
| GPT-5 | 2.98 | 7.32 | -4.33s (-59.2%) |
| Gemini-2.5-flash | 457.36 | 554.35 | -96.99s (-17.5%) |
| Gemini-2.5-flash-nothinking | 168.14 | 132.33 | +35.81s (+27.1%) |
| Qwen3-235b | 10.49 | 9.54 | +0.95s (+10.0%) |
GameBuilder: MCP vs Hardcoded
Overall Summary (Averaged Across All Models)
| MCP Mean (s) | Hardcoded Mean (s) | Diff (MCP-Hard) |
|---|---|---|
| 306.74 | 317.61 | -10.87s (-3.4%) |
Per-Model Comparison
| Model | MCP Mean (s) | Hardcoded Mean (s) | Diff (MCP-Hard) |
|---|---|---|---|
| DeepSeek-R1 | 827.60 | 914.94 | -87.34s (-9.5%) |
| DeepSeek-V3-1 | 516.71 | 518.26 | -1.55s (-0.3%) |
| GPT-4o-mini | 136.40 | 94.49 | +41.90s (+44.3%) |
| GPT-5 | 43.16 | 45.35 | -2.20s (-4.8%) |
| Gemini-2.5-flash | 185.74 | 163.51 | +22.23s (+13.6%) |
| Gemini-2.5-flash-nothinking | 120.75 | 169.19 | -48.44s (-28.6%) |
| Qwen3-235b | 316.86 | 317.52 | -0.66s (-0.2%) |
EmailResponder: MCP vs Hardcoded
Overall Summary (Averaged Across All Models)
| MCP Mean (s) | Hardcoded Mean (s) | Diff (MCP-Hard) |
|---|---|---|
| 73.97 | 105.83 | -31.86s (-30.1%) |
Per-Model Comparison
| Model | MCP Mean (s) | Hardcoded Mean (s) | Diff (MCP-Hard) |
|---|---|---|---|
| DeepSeek-R1 | 141.80 | 233.49 | -91.69s (-39.3%) |
| DeepSeek-V3-1 | 44.42 | 47.23 | -2.81s (-6.0%) |
| GPT-4o-mini | 18.15 | 17.90 | +0.25s (+1.4%) |
| GPT-5 | 8.25 | 9.41 | -1.16s (-12.3%) |
| Gemini-2.5-flash | 115.85 | 180.03 | -64.18s (-35.6%) |
| Gemini-2.5-flash-nothinking | 151.00 | 217.85 | -66.85s (-30.7%) |
| Qwen3-235b | 38.30 | 37.41 | +0.89s (+2.4%) |