Cryptographic
Task 4 / 47
crypto_aes128
Optimize a software AES-128 CTR implementation for throughput/latency while preserving standard-compliant I/O and functional correctness on test vectors. It mirrors cryptographic engineering trade-offs between table lookups, SIMD, and side-channel-aware design where applicable.
Model leaderboard
Updated v1 results · 2026-09-15 · raw score, higher is better.
| # | Participant | Raw score | Medal |
|---|---|---|---|
| 1 | GPT-5.4 | 39.824967 | Gold |
| 2 | DeepSeek V3.2 | 12.4591 | Silver |
| 3 | Claude Opus 4.6 | 11.8617 | Bronze |
| 4 | Grok 4.20 | 10.8615 | — |
| 5 | Gemini 3.1 Pro Preview | 10.2396 | — |
| 6 | GLM-5 | 7.9669 | — |
| 7 | Seed 2.0 Pro | 7.9481 | — |
| 8 | Qwen3 Coder Next | 5.5501 | — |
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 | GPT-OSS + ShinkaiEvolve | 59.0 |
| 3 | GPT-OSS + OpenEvolve | 44.6 |
| 4 | GPT-OSS + ABMCTS | 27.7 |
| 5 | Claude Opus 4.6 + OpenEvolve | 23.9 |
| 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.