Computer Systems
Task 3 / 47
MallocLab
The CSAPP-style Malloc Lab: implement `malloc`/`free`/`realloc` in a constrained region to maximize utilization and throughput on realistic allocation traces. It exercises segregated lists, coalescing, and throughput-oriented heuristics, scored by an mdriver-like harness for correctness and performance.
Model leaderboard
Updated v1 results · 2026-09-15 · raw score, higher is better.
| # | Participant | Raw score | Medal |
|---|---|---|---|
| 1 | Claude Opus 4.6 | 84.055903 | Gold |
| 2 | GLM-5 | 68.137464 | Silver |
| 3 | Gemini 3.1 Pro Preview | 50.188204 | Bronze |
| 4 | Grok 4.20 | 50.064197 | — |
| 5 | DeepSeek V3.2 | 49.254088 | — |
| 6 | Seed 2.0 Pro | 30.74004 | — |
| 7 | GPT-5.4 | 28.145173 | — |
| — | Qwen3 Coder Next | No valid score | — |
Framework results · original paper
Historical results from the original evaluators · normalized score (0–100).
| # | Participant | Score |
|---|---|---|
| 1 | Claude Opus 4.6 + ShinkaiEvolve | 100.0 |
| 2 | Claude Opus 4.6 + OpenEvolve | 98.6 |
| 3 | GPT-OSS + ShinkaiEvolve | 84.1 |
| 4 | GPT-OSS + OpenEvolve | 36.2 |
| 5 | GPT-OSS + ABMCTS | 29.0 |
| 6 | Claude Opus 4.6 + ABMCTS | 0.0 |
Model results use the updated scores and medal thresholds. No valid score earns zero medal credit. Historical framework scores use the original evaluators and have not been updated alongside the model results.