# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260402-142215
**Date:** Apr 02, 2026
**Category:** code
**Question ID:** CODE-023

---

## Question

Write a database migration that adds a NOT NULL column with a default value to a table with 50 million rows in PostgreSQL — without locking the table or causing downtime. Explain the step-by-step process, why naive ALTER TABLE fails at scale, and how to verify data integrity after migration.

---

## Winner

**GPT-5.4** (openrouter)
- Winner Score: 8.79
- Matrix Average: 7.70
- Total Judgments: 84

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | GPT-5.4 | openrouter | 8.79 | 9 |
| 2 | Grok 4.20 | openrouter | 8.40 | 9 |
| 3 | Gemini 3 Flash Preview | Google | 8.27 | 9 |
| 4 | Claude Sonnet 4.6 | openrouter | 8.08 | 9 |
| 5 | MiMo-V2-Flash | Xiaomi | 7.96 | 9 |
| 6 | Claude Opus 4.6 | openrouter | 7.74 | 9 |
| 7 | GPT-OSS-120B | OpenAI | 7.39 | 9 |
| 8 | DeepSeek V4 | openrouter | 6.95 | 8 |
| 9 | MiniMax M2.5 | openrouter | 6.76 | 4 |
| 10 | Gemini 3.1 Pro | openrouter | 6.66 | 9 |

---

## 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 | — | 4.5 | 4.3 | 5.8 | 7.5 | · | 5.8 | 7.7 | · | 6.3 |
| Claude Opus | 9.3 | — | 6.5 | 7.0 | 8.0 | 5.6 | 6.2 | 8.0 | 0.5 | 8.0 |
| Gemini 3.1 Pro | 6.0 | 4.7 | — | 5.7 | 5.4 | 4.2 | 3.5 | 6.7 | · | 7.2 |
| Claude Sonnet | 9.0 | 8.6 | 6.3 | — | 8.8 | 7.3 | 8.0 | 8.6 | · | 8.3 |
| Grok 4.20 | 8.8 | 8.4 | 6.6 | 8.7 | — | 6.6 | 8.6 | 8.0 | 8.6 | 6.0 |
| DeepSeek V4 | 10.0 | 10.0 | 8.6 | 9.6 | 9.6 | — | 9.3 | 9.6 | 8.6 | 9.6 |
| GPT-OSS-120B | 8.8 | 7.3 | 5.3 | 8.8 | 8.8 | 7.8 | — | 8.6 | · | 8.6 |
| Gemini 3 | 9.6 | 9.6 | 7.8 | 10.0 | 9.6 | 7.5 | 9.2 | — | · | 9.6 |
| MiniMax M2.5 | 8.6 | 7.6 | 5.8 | 7.8 | 9.0 | 8.0 | 7.3 | 8.6 | — | 8.1 |
| MiMo-V2-Flash | 9.0 | 9.0 | 8.8 | 9.3 | 9.0 | 8.6 | 8.6 | 8.8 | 9.3 | — |

---

## 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-023. 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-20260402-142215/results
Full dataset: https://app.themultivac.com/dashboard/export