# Overall MCP vs A2A Comparison

Averaged across all projects: BookWriter, LandingPageGenerator, RecruitmentAssistant, SQLAssistant, SocialMediaManager

## Framework-Level Comparison (All Models Averaged)

| Metric | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| exact_match | 0.07 | 0.01 | +0.06 |
| any_order_match | 0.29 | 0.25 | +0.04 |
| precision | 0.85 | 0.82 | +0.03 |
| recall | 0.81 | 0.84 | -0.03 |
| retry_rate | 0.21 | 0.22 | -0.02 |
| average_score | 0.47 | 0.31 | +0.17 |
| pass_rate | 0.99 | 0.99 | +0.00 |

---

## exact_match

| Model | MCP (Avg) | A2A (Avg) | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.00 | 0.00 | +0.00 |
| DeepSeek-V3-1 | 0.20 | 0.03 | +0.17 |
| GPT-4o-mini | 0.04 | 0.01 | +0.03 |
| GPT-5 | 0.18 | 0.01 | +0.17 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.07 | 0.01 | +0.06 |

## any_order_match

| Model | MCP (Avg) | A2A (Avg) | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.14 | 0.05 | +0.10 |
| DeepSeek-V3-1 | 0.63 | 0.48 | +0.15 |
| GPT-4o-mini | 0.54 | 0.60 | -0.06 |
| GPT-5 | 0.29 | 0.17 | +0.12 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.42 | 0.43 | -0.01 |

## precision

| Model | MCP (Avg) | A2A (Avg) | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.71 | 0.70 | +0.02 |
| DeepSeek-V3-1 | 0.84 | 0.74 | +0.10 |
| GPT-4o-mini | 0.72 | 0.79 | -0.06 |
| GPT-5 | 0.90 | 0.89 | +0.02 |
| Gemini-2.5-flash | 0.97 | 0.91 | +0.06 |
| Gemini-2.5-flash-nothinking | 0.97 | 0.93 | +0.04 |
| Qwen3-235b | 0.84 | 0.81 | +0.03 |

## recall

| Model | MCP (Avg) | A2A (Avg) | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.88 | 0.88 | -0.00 |
| DeepSeek-V3-1 | 0.98 | 0.95 | +0.02 |
| GPT-4o-mini | 0.95 | 0.96 | -0.01 |
| GPT-5 | 0.88 | 0.90 | -0.02 |
| Gemini-2.5-flash | 0.55 | 0.62 | -0.07 |
| Gemini-2.5-flash-nothinking | 0.57 | 0.64 | -0.07 |
| Qwen3-235b | 0.90 | 0.93 | -0.03 |

## retry_rate

| Model | MCP (Avg) | A2A (Avg) | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.35 | 0.37 | -0.02 |
| DeepSeek-V3-1 | 0.19 | 0.25 | -0.06 |
| GPT-4o-mini | 0.26 | 0.15 | +0.11 |
| GPT-5 | 0.23 | 0.18 | +0.05 |
| Gemini-2.5-flash | 0.09 | 0.19 | -0.10 |
| Gemini-2.5-flash-nothinking | 0.09 | 0.18 | -0.09 |
| Qwen3-235b | 0.23 | 0.25 | -0.02 |

## average_score

| Model | MCP (Avg) | A2A (Avg) | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.51 | 0.16 | +0.34 |
| DeepSeek-V3-1 | 0.66 | 0.49 | +0.17 |
| GPT-4o-mini | 0.63 | 0.57 | +0.06 |
| GPT-5 | 0.63 | 0.33 | +0.30 |
| Gemini-2.5-flash | 0.09 | 0.08 | +0.01 |
| Gemini-2.5-flash-nothinking | 0.10 | 0.03 | +0.08 |
| Qwen3-235b | 0.70 | 0.49 | +0.21 |

## pass_rate

| Model | MCP (Avg) | A2A (Avg) | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.98 | 0.97 | +0.02 |
| DeepSeek-V3-1 | 1.00 | 1.00 | +0.00 |
| GPT-4o-mini | 0.98 | 0.98 | +0.01 |
| GPT-5 | 0.99 | 0.99 | +0.01 |
| Gemini-2.5-flash | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 1.00 | 1.00 | +0.00 |
| Qwen3-235b | 0.98 | 0.99 | -0.01 |

---

# BookWriter: MCP vs A2A

## Overall Summary (Averaged Across All Models)

| Metric | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| exact_match | 0.02 | 0.00 | +0.02 |
| any_order_match | 0.09 | 0.08 | +0.01 |
| precision | 0.94 | 0.93 | +0.01 |
| recall | 0.73 | 0.77 | -0.04 |
| retry_rate | 0.02 | 0.02 | +0.00 |
| pass_rate | 1.00 | 1.00 | +0.00 |

---

## exact_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.00 | 0.00 | +0.00 |
| DeepSeek-V3-1 | 0.13 | 0.00 | +0.13 |
| GPT-4o-mini | 0.00 | 0.00 | +0.00 |
| GPT-5 | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.00 | 0.00 | +0.00 |

## any_order_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.02 | 0.02 | +0.00 |
| DeepSeek-V3-1 | 0.63 | 0.53 | +0.10 |
| GPT-4o-mini | 0.00 | 0.00 | +0.00 |
| GPT-5 | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.00 | 0.00 | +0.00 |

## precision

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.86 | 0.87 | -0.02 |
| DeepSeek-V3-1 | 0.96 | 0.79 | +0.17 |
| GPT-4o-mini | 0.84 | 0.90 | -0.06 |
| GPT-5 | 0.98 | 0.98 | +0.01 |
| Gemini-2.5-flash | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 1.00 | 1.00 | -0.00 |
| Qwen3-235b | 0.96 | 0.96 | -0.00 |

## recall

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.80 | 0.84 | -0.04 |
| DeepSeek-V3-1 | 0.99 | 0.99 | -0.00 |
| GPT-4o-mini | 0.85 | 0.86 | -0.01 |
| GPT-5 | 0.81 | 0.82 | -0.02 |
| Gemini-2.5-flash | 0.47 | 0.53 | -0.06 |
| Gemini-2.5-flash-nothinking | 0.47 | 0.53 | -0.06 |
| Qwen3-235b | 0.73 | 0.80 | -0.06 |

## retry_rate

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.13 | 0.13 | +0.00 |
| DeepSeek-V3-1 | 0.00 | 0.02 | -0.02 |
| GPT-4o-mini | 0.00 | 0.00 | +0.00 |
| GPT-5 | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.02 | 0.00 | +0.02 |
| Qwen3-235b | 0.02 | 0.00 | +0.02 |

## pass_rate

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 1.00 | 1.00 | +0.00 |
| DeepSeek-V3-1 | 1.00 | 1.00 | +0.00 |
| GPT-4o-mini | 1.00 | 1.00 | +0.00 |
| GPT-5 | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 1.00 | 1.00 | +0.00 |
| Qwen3-235b | 1.00 | 1.00 | +0.00 |

# LandingPageGenerator: MCP vs A2A

## Overall Summary (Averaged Across All Models)

| Metric | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| exact_match | 0.29 | 0.00 | +0.29 |
| any_order_match | 0.60 | 0.40 | +0.20 |
| precision | 0.92 | 0.84 | +0.08 |
| recall | 0.90 | 0.91 | -0.01 |
| retry_rate | 0.03 | 0.02 | +0.01 |
| average_score | 0.57 | 0.30 | +0.27 |

---

## exact_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.00 | 0.00 | +0.00 |
| DeepSeek-V3-1 | 0.72 | 0.00 | +0.72 |
| GPT-4o-mini | 0.12 | 0.00 | +0.12 |
| GPT-5 | 0.88 | 0.00 | +0.88 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.28 | 0.00 | +0.28 |

## any_order_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.32 | 0.05 | +0.27 |
| DeepSeek-V3-1 | 1.00 | 1.00 | +0.00 |
| GPT-4o-mini | 0.98 | 0.97 | +0.02 |
| GPT-5 | 1.00 | 0.27 | +0.73 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.87 | 0.48 | +0.38 |

## precision

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.80 | 0.76 | +0.04 |
| DeepSeek-V3-1 | 0.97 | 0.83 | +0.13 |
| GPT-4o-mini | 0.82 | 0.74 | +0.08 |
| GPT-5 | 0.99 | 0.81 | +0.18 |
| Gemini-2.5-flash | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.99 | 0.98 | +0.00 |
| Qwen3-235b | 0.91 | 0.79 | +0.12 |

## recall

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.94 | 0.94 | +0.00 |
| DeepSeek-V3-1 | 1.00 | 1.00 | +0.00 |
| GPT-4o-mini | 1.00 | 1.00 | +0.00 |
| GPT-5 | 1.00 | 0.97 | +0.03 |
| Gemini-2.5-flash | 0.67 | 0.74 | -0.07 |
| Gemini-2.5-flash-nothinking | 0.70 | 0.76 | -0.07 |
| Qwen3-235b | 0.99 | 0.98 | +0.01 |

## retry_rate

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.15 | 0.03 | +0.12 |
| DeepSeek-V3-1 | 0.00 | 0.00 | +0.00 |
| GPT-4o-mini | 0.00 | 0.00 | +0.00 |
| GPT-5 | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.03 | 0.05 | -0.02 |
| Qwen3-235b | 0.00 | 0.05 | -0.05 |

## average_score

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.58 | 0.08 | +0.50 |
| DeepSeek-V3-1 | 0.83 | 0.86 | -0.02 |
| GPT-4o-mini | 0.86 | 0.57 | +0.30 |
| GPT-5 | 0.91 | 0.19 | +0.71 |
| Gemini-2.5-flash | 0.01 | 0.00 | +0.01 |
| Gemini-2.5-flash-nothinking | 0.03 | 0.00 | +0.03 |
| Qwen3-235b | 0.78 | 0.38 | +0.40 |

# RecruitmentAssistant: MCP vs A2A

## Overall Summary (Averaged Across All Models)

| Metric | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| exact_match | 0.01 | 0.00 | +0.00 |
| any_order_match | 0.21 | 0.18 | +0.03 |
| precision | 0.87 | 0.89 | -0.02 |
| recall | 0.83 | 0.86 | -0.03 |
| retry_rate | 0.05 | 0.05 | +0.00 |
| pass_rate | 1.00 | 1.00 | +0.00 |

---

## exact_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.00 | 0.00 | +0.00 |
| DeepSeek-V3-1 | 0.00 | 0.00 | +0.00 |
| GPT-4o-mini | 0.00 | 0.00 | +0.00 |
| GPT-5 | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.05 | 0.02 | +0.03 |

## any_order_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.08 | 0.00 | +0.08 |
| DeepSeek-V3-1 | 0.07 | 0.00 | +0.07 |
| GPT-4o-mini | 0.70 | 0.68 | +0.02 |
| GPT-5 | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.60 | 0.57 | +0.03 |

## precision

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.79 | 0.83 | -0.04 |
| DeepSeek-V3-1 | 0.84 | 0.87 | -0.03 |
| GPT-4o-mini | 0.74 | 0.79 | -0.05 |
| GPT-5 | 0.99 | 0.99 | -0.00 |
| Gemini-2.5-flash | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.99 | 1.00 | -0.00 |
| Qwen3-235b | 0.73 | 0.74 | -0.01 |

## recall

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.88 | 0.88 | +0.00 |
| DeepSeek-V3-1 | 0.94 | 0.96 | -0.02 |
| GPT-4o-mini | 0.98 | 0.98 | -0.01 |
| GPT-5 | 0.89 | 0.90 | -0.01 |
| Gemini-2.5-flash | 0.55 | 0.64 | -0.09 |
| Gemini-2.5-flash-nothinking | 0.60 | 0.68 | -0.08 |
| Qwen3-235b | 0.96 | 0.97 | -0.02 |

## retry_rate

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.03 | 0.07 | -0.03 |
| DeepSeek-V3-1 | 0.00 | 0.00 | +0.00 |
| GPT-4o-mini | 0.00 | 0.00 | +0.00 |
| GPT-5 | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.33 | 0.27 | +0.07 |

## pass_rate

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 1.00 | 0.98 | +0.02 |
| DeepSeek-V3-1 | 1.00 | 1.00 | +0.00 |
| GPT-4o-mini | 1.00 | 1.00 | +0.00 |
| GPT-5 | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 1.00 | 1.00 | +0.00 |
| Qwen3-235b | 0.97 | 0.98 | -0.02 |

# SQLAssistant: MCP vs A2A

## Overall Summary (Averaged Across All Models)

| Metric | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| exact_match | 0.00 | 0.00 | +0.00 |
| any_order_match | 0.14 | 0.15 | -0.02 |
| precision | 0.72 | 0.65 | +0.07 |
| recall | 0.86 | 0.87 | -0.01 |
| retry_rate | 0.53 | 0.65 | -0.12 |
| average_score | 0.38 | 0.32 | +0.06 |

---

## exact_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.00 | 0.00 | +0.00 |
| DeepSeek-V3-1 | 0.00 | 0.00 | +0.00 |
| GPT-4o-mini | 0.00 | 0.00 | +0.00 |
| GPT-5 | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.00 | 0.00 | +0.00 |

## any_order_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.10 | 0.03 | +0.07 |
| DeepSeek-V3-1 | 0.52 | 0.00 | +0.52 |
| GPT-4o-mini | 0.12 | 0.53 | -0.42 |
| GPT-5 | 0.00 | 0.03 | -0.03 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.23 | 0.48 | -0.25 |

## precision

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.50 | 0.41 | +0.10 |
| DeepSeek-V3-1 | 0.67 | 0.41 | +0.26 |
| GPT-4o-mini | 0.61 | 0.84 | -0.23 |
| GPT-5 | 0.71 | 0.84 | -0.13 |
| Gemini-2.5-flash | 0.89 | 0.57 | +0.32 |
| Gemini-2.5-flash-nothinking | 0.90 | 0.70 | +0.20 |
| Qwen3-235b | 0.78 | 0.78 | -0.00 |

## recall

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.90 | 0.85 | +0.05 |
| DeepSeek-V3-1 | 0.96 | 0.84 | +0.12 |
| GPT-4o-mini | 0.92 | 0.98 | -0.06 |
| GPT-5 | 0.85 | 0.88 | -0.04 |
| Gemini-2.5-flash | 0.71 | 0.77 | -0.07 |
| Gemini-2.5-flash-nothinking | 0.71 | 0.78 | -0.07 |
| Qwen3-235b | 0.96 | 0.96 | -0.00 |

## retry_rate

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.78 | 0.95 | -0.17 |
| DeepSeek-V3-1 | 0.57 | 1.00 | -0.43 |
| GPT-4o-mini | 0.65 | 0.17 | +0.48 |
| GPT-5 | 0.68 | 0.40 | +0.28 |
| Gemini-2.5-flash | 0.30 | 0.90 | -0.60 |
| Gemini-2.5-flash-nothinking | 0.32 | 0.75 | -0.43 |
| Qwen3-235b | 0.38 | 0.38 | +0.00 |

## average_score

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.44 | 0.25 | +0.19 |
| DeepSeek-V3-1 | 0.48 | 0.12 | +0.36 |
| GPT-4o-mini | 0.40 | 0.58 | -0.18 |
| GPT-5 | 0.36 | 0.47 | -0.11 |
| Gemini-2.5-flash | 0.16 | 0.16 | +0.01 |
| Gemini-2.5-flash-nothinking | 0.18 | 0.05 | +0.13 |
| Qwen3-235b | 0.62 | 0.60 | +0.02 |

# SocialMediaManager: MCP vs A2A

## Overall Summary (Averaged Across All Models)

| Metric | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| exact_match | 0.03 | 0.04 | -0.01 |
| any_order_match | 0.40 | 0.42 | -0.01 |
| precision | 0.80 | 0.80 | -0.00 |
| recall | 0.75 | 0.79 | -0.04 |
| retry_rate | 0.40 | 0.39 | +0.01 |
| pass_rate | 0.98 | 0.97 | +0.01 |

---

## exact_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.00 | 0.00 | +0.00 |
| DeepSeek-V3-1 | 0.15 | 0.15 | +0.00 |
| GPT-4o-mini | 0.07 | 0.03 | +0.03 |
| GPT-5 | 0.00 | 0.05 | -0.05 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.00 | 0.03 | -0.03 |

## any_order_match

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.20 | 0.13 | +0.07 |
| DeepSeek-V3-1 | 0.93 | 0.85 | +0.08 |
| GPT-4o-mini | 0.88 | 0.80 | +0.08 |
| GPT-5 | 0.43 | 0.53 | -0.10 |
| Gemini-2.5-flash | 0.00 | 0.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 0.00 | 0.00 | +0.00 |
| Qwen3-235b | 0.38 | 0.62 | -0.23 |

## precision

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.62 | 0.60 | +0.02 |
| DeepSeek-V3-1 | 0.74 | 0.79 | -0.05 |
| GPT-4o-mini | 0.61 | 0.66 | -0.04 |
| GPT-5 | 0.85 | 0.83 | +0.02 |
| Gemini-2.5-flash | 0.97 | 0.98 | -0.02 |
| Gemini-2.5-flash-nothinking | 0.98 | 0.97 | +0.01 |
| Qwen3-235b | 0.85 | 0.80 | +0.05 |

## recall

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.86 | 0.89 | -0.02 |
| DeepSeek-V3-1 | 0.98 | 0.98 | +0.01 |
| GPT-4o-mini | 0.99 | 0.99 | +0.00 |
| GPT-5 | 0.86 | 0.90 | -0.04 |
| Gemini-2.5-flash | 0.35 | 0.44 | -0.09 |
| Gemini-2.5-flash-nothinking | 0.35 | 0.43 | -0.09 |
| Qwen3-235b | 0.87 | 0.93 | -0.06 |

## retry_rate

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.65 | 0.65 | +0.00 |
| DeepSeek-V3-1 | 0.38 | 0.23 | +0.15 |
| GPT-4o-mini | 0.67 | 0.60 | +0.07 |
| GPT-5 | 0.45 | 0.48 | -0.03 |
| Gemini-2.5-flash | 0.15 | 0.07 | +0.08 |
| Gemini-2.5-flash-nothinking | 0.08 | 0.12 | -0.03 |
| Qwen3-235b | 0.42 | 0.55 | -0.13 |

## pass_rate

| Model | MCP | A2A | Diff (MCP-A2A) |
| --- | --- | --- | --- |
| DeepSeek-R1 | 0.95 | 0.92 | +0.03 |
| DeepSeek-V3-1 | 1.00 | 1.00 | +0.00 |
| GPT-4o-mini | 0.95 | 0.93 | +0.02 |
| GPT-5 | 0.98 | 0.97 | +0.02 |
| Gemini-2.5-flash | 1.00 | 1.00 | +0.00 |
| Gemini-2.5-flash-nothinking | 1.00 | 1.00 | +0.00 |
| Qwen3-235b | 0.98 | 0.98 | +0.00 |