# The Multivac — Evaluation Report

**Evaluation ID:** EVAL-20260402-143013
**Date:** Apr 02, 2026
**Category:** code
**Question ID:** CODE-024

---

## Question

Implement an HTTP/1.1 server from raw TCP sockets in Python (no http.server, no frameworks). Support: GET and POST methods, proper header parsing, chunked transfer encoding, keep-alive connections, static file serving, and a 404 handler. Handle malformed requests gracefully.

---

## Winner

**MiniMax M2.5** (openrouter)
- Winner Score: 8.29
- Matrix Average: 6.61
- Total Judgments: 86

---

## Rankings

| Rank | Model | Provider | Avg Score | Judgments |
|------|-------|----------|-----------|----------|
| 1 | MiniMax M2.5 | openrouter | 8.29 | 9 |
| 2 | Gemini 3 Flash Preview | Google | 7.69 | 9 |
| 3 | GPT-5.4 | openrouter | 7.41 | 8 |
| 4 | Grok 4.20 | openrouter | 7.30 | 9 |
| 5 | DeepSeek V4 | openrouter | 7.15 | 8 |
| 6 | Claude Sonnet 4.6 | openrouter | 6.61 | 9 |
| 7 | GPT-OSS-120B | OpenAI | 6.48 | 9 |
| 8 | MiMo-V2-Flash | Xiaomi | 5.96 | 8 |
| 9 | Claude Opus 4.6 | openrouter | 5.63 | 9 |
| 10 | Gemini 3.1 Pro | openrouter | 3.53 | 8 |

---

## 10×10 Judgment Matrix

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

| Judge ↓ / Resp → | Gemini 3.1 Pro | GPT-5.4 | Grok 4.20 | Claude Opus | Claude Sonnet | DeepSeek V4 | GPT-OSS-120B | Gemini 3 | MiniMax M2.5 | MiMo-V2-Flash |
|---|---|---|---|---|---|---|---|---|---|---|
| Gemini 3.1 Pro | — | 6.0 | 7.0 | 4.9 | 4.1 | 5.4 | 4.1 | 7.6 | 9.6 | 4.3 |
| GPT-5.4 | 1.6 | — | 5.0 | 2.3 | 2.6 | 6.2 | 3.3 | 6.8 | 6.8 | 3.3 |
| Grok 4.20 | 3.6 | 6.6 | — | 6.0 | 8.7 | 6.8 | 5.8 | 6.8 | 6.8 | 6.0 |
| Claude Opus | 1.2 | 7.4 | 6.4 | — | 6.5 | 6.4 | 5.3 | 7.4 | 7.5 | · |
| Claude Sonnet | 1.9 | 8.2 | 7.0 | 6.0 | — | 7.0 | 7.2 | 7.8 | 8.2 | 6.8 |
| DeepSeek V4 | 8.2 | 9.0 | 8.6 | 8.0 | 8.6 | — | 8.8 | 8.6 | 8.8 | 8.8 |
| GPT-OSS-120B | 1.1 | 5.0 | 7.3 | 4.3 | 6.0 | 8.6 | — | 8.0 | 8.6 | 4.5 |
| Gemini 3 | 2.9 | 9.2 | 9.2 | 7.5 | 8.8 | 9.2 | 8.4 | — | 9.6 | 8.6 |
| MiniMax M2.5 | · | 8.1 | 7.2 | 5.8 | 5.5 | · | 6.8 | 8.2 | — | 5.4 |
| MiMo-V2-Flash | 7.7 | · | 8.2 | 5.8 | 8.8 | 7.8 | 8.6 | 8.2 | 8.9 | — |

---

## 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-024. 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-20260402-143013/results
Full dataset: https://app.themultivac.com/dashboard/export