# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260402-121140
**Date:** Jan 13, 2026
**Category:** code
**Question ID:** CODE-001

---

## Question

This Python async function has 3 bugs: a race condition, an unhandled exception, and a resource leak. Find all three and explain why each is problematic.

```python
import asyncio
import aiohttp

class DataFetcher:
    def __init__(self):
        self.cache = {}
        self.session = aiohttp.ClientSession()
    
    async def fetch_data(self, urls):
        results = []
        for url in urls:
            if url in self.cache:
                results.append(self.cache[url])
            else:
                async with self.session.get(url) as response:
                    data = await response.json()
                    self.cache[url] = data
                    results.append(data)
        return results
    
    async def fetch_parallel(self, urls):
        tasks = [self.fetch_single(url) for url in urls]
        return await asyncio.gather(*tasks)
    
    async def fetch_single(self, url):
        if url in self.cache:
            return self.cache[url]
        async with self.session.get(url) as response:
            data = await response.json()
            self.cache[url] = data
            return data
```

---

## Winner

**Grok 4.20** (openrouter)
- Winner Score: 9.44
- Matrix Average: 8.51
- Total Judgments: 86

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | Grok 4.20 | openrouter | 9.44 | 9 |
| 2 | Claude Opus 4.6 | openrouter | 9.33 | 9 |
| 3 | Claude Sonnet 4.6 | openrouter | 9.32 | 9 |
| 4 | Gemini 3 Flash Preview | Google | 9.27 | 9 |
| 5 | GPT-5.4 | openrouter | 9.13 | 9 |
| 6 | MiMo-V2-Flash | Xiaomi | 8.93 | 9 |
| 7 | Gemini 3.1 Pro | openrouter | 8.76 | 9 |
| 8 | DeepSeek V4 | openrouter | 8.68 | 9 |
| 9 | MiniMax M2.5 | openrouter | 7.93 | 9 |
| 10 | GPT-OSS-120B | OpenAI | 4.31 | 5 |

---

## 10×10 Judgment Matrix

Rows = Judge, Columns = Respondent. Self-judgments excluded (—).

| Judge ↓ / Resp → | GPT-5.4 | Claude Opus | Gemini 3.1 Pro | Claude Sonnet | Grok 4.20 | DeepSeek V4 | GPT-OSS-120B | Gemini 3 | MiniMax M2.5 | MiMo-V2-Flash |
|---|---|---|---|---|---|---|---|---|---|---|
| GPT-5.4 | — | 8.8 | 8.3 | 9.0 | 9.0 | 7.8 | 0.3 | 8.6 | 6.7 | 8.3 |
| Claude Opus | 9.2 | — | 8.2 | 9.6 | 8.9 | 8.1 | 0.7 | 9.2 | 7.5 | 9.2 |
| Gemini 3.1 Pro | 9.7 | 10.0 | — | 10.0 | 10.0 | 8.0 | · | 10.0 | 6.3 | 8.5 |
| Claude Sonnet | 9.0 | 9.3 | 8.2 | — | 9.0 | 8.0 | · | 8.8 | 8.3 | 8.8 |
| Grok 4.20 | 8.6 | 8.3 | 8.6 | 8.0 | — | 8.6 | 4.2 | 9.0 | 8.6 | 6.8 |
| DeepSeek V4 | 9.4 | 8.8 | 9.4 | 10.0 | 9.6 | — | 7.6 | 9.6 | 9.6 | 10.0 |
| GPT-OSS-120B | 8.4 | 9.3 | 8.1 | 8.8 | 9.3 | 8.8 | — | 8.6 | 7.3 | 8.8 |
| Gemini 3 | 9.8 | 10.0 | 9.6 | 10.0 | 9.8 | 9.8 | · | — | 8.6 | 10.0 |
| MiniMax M2.5 | 8.8 | 9.4 | 8.4 | 9.0 | 9.6 | 10.0 | · | 9.8 | — | 10.0 |
| MiMo-V2-Flash | 9.2 | 10.0 | 10.0 | 9.6 | 9.8 | 9.0 | 8.8 | 9.8 | 8.6 | — |

---

## Methodology

- **10×10 Blind Peer Matrix:** All models answer the same question, then all models judge all responses.
- **5 Criteria:** Correctness, completeness, clarity, depth, usefulness (each scored 1–10).
- **Self-judgments excluded:** Models do not judge their own responses.
- **Weighted Score:** Composite of all 5 criteria.

---

## Citation

The Multivac (2026). Blind Peer Evaluation: CODE-001. app.themultivac.com

## License

Open data. Free to use, share, and build upon. Please cite The Multivac when using this data.

Download raw JSON: https://app.themultivac.com/api/evaluations/EVAL-20260402-121140/results
Full dataset: https://app.themultivac.com/dashboard/export