# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260402-124935
**Date:** Mar 10, 2026
**Category:** code
**Question ID:** CODE-009

---

## Question

This Python application has a memory leak. Find it and explain the fix.

```python
import threading
import time
from functools import lru_cache

class EventProcessor:
    _instances = []
    
    def __init__(self, name):
        self.name = name
        self.callbacks = []
        self._lock = threading.Lock()
        EventProcessor._instances.append(self)
    
    def register_callback(self, func):
        self.callbacks.append(func)
    
    @lru_cache(maxsize=10000)
    def process(self, event_data):
        results = []
        for callback in self.callbacks:
            result = callback(event_data)
            results.append(result)
        return tuple(results)
    
    def __del__(self):
        print(f"Processor {self.name} deleted")

def create_processor_for_request(request_id):
    processor = EventProcessor(f"processor_{request_id}")
    processor.register_callback(lambda x: x.upper())
    processor.register_callback(lambda x: len(x))
    return processor

# Simulated request handling
def handle_request(request_id, data):
    processor = create_processor_for_request(request_id)
    return processor.process(data)

# This runs for hours...
for i in range(1000000):
    result = handle_request(i, f"event_data_{i}")
    time.sleep(0.001)
```

---

## Winner

**GPT-5.4** (openrouter)
- Winner Score: 9.54
- Matrix Average: 8.07
- Total Judgments: 89

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | GPT-5.4 | openrouter | 9.54 | 9 |
| 2 | Grok 4.20 | openrouter | 9.06 | 9 |
| 3 | Claude Opus 4.6 | openrouter | 8.81 | 9 |
| 4 | Claude Sonnet 4.6 | openrouter | 8.69 | 8 |
| 5 | GPT-OSS-120B | OpenAI | 8.46 | 9 |
| 6 | Gemini 3 Flash Preview | Google | 8.46 | 9 |
| 7 | DeepSeek V4 | openrouter | 7.55 | 9 |
| 8 | MiniMax M2.5 | openrouter | 7.11 | 9 |
| 9 | Gemini 3.1 Pro | openrouter | 6.62 | 9 |
| 10 | MiMo-V2-Flash | Xiaomi | 6.42 | 9 |

---

## 10×10 Judgment Matrix

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

| Judge ↓ / Resp → | MiniMax M2.5 | Gemini 3.1 Pro | Grok 4.20 | GPT-5.4 | Claude Opus | Claude Sonnet | DeepSeek V4 | GPT-OSS-120B | Gemini 3 | MiMo-V2-Flash |
|---|---|---|---|---|---|---|---|---|---|---|
| MiniMax M2.5 | — | 6.8 | 8.4 | 10.0 | 8.8 | 8.6 | 8.6 | 8.6 | 8.0 | 8.0 |
| Gemini 3.1 Pro | 4.8 | — | 8.5 | 10.0 | 8.8 | 6.8 | 7.0 | 6.1 | 8.5 | 3.6 |
| Grok 4.20 | 6.5 | 5.8 | — | 9.0 | 7.5 | · | 6.6 | 8.8 | 7.8 | 6.0 |
| GPT-5.4 | 5.8 | 3.3 | 8.6 | — | 8.0 | 8.0 | 6.5 | 5.3 | 8.2 | 4.8 |
| Claude Opus | 5.0 | 5.3 | 9.2 | 9.8 | — | 8.9 | 6.3 | 9.3 | 8.2 | 5.5 |
| Claude Sonnet | 6.8 | 5.9 | 9.0 | 9.3 | 8.8 | — | 7.8 | 9.2 | 8.4 | 6.0 |
| DeepSeek V4 | 8.8 | 9.3 | 9.4 | 9.6 | 9.6 | 9.8 | — | 9.6 | 9.6 | 9.6 |
| GPT-OSS-120B | 8.2 | 5.5 | 8.6 | 8.8 | 8.6 | 8.1 | 7.5 | — | 8.6 | 7.5 |
| Gemini 3 | 9.6 | 9.3 | 10.0 | 10.0 | 10.0 | 10.0 | 9.2 | 9.8 | — | 6.8 |
| MiMo-V2-Flash | 8.6 | 8.4 | 9.8 | 9.3 | 9.2 | 9.3 | 8.6 | 9.3 | 8.8 | — |

---

## 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-009. 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-20260402-124935/results
Full dataset: https://app.themultivac.com/dashboard/export