Synodic-AI-Sovereign
High-Precision Code Synthesis & Reasoning Model
Developed by Synodic-AI Corp.
📊 Benchmark Results
Evaluated on standard benchmark sets using LiveCodeBench, EvalPlus v0.8.0, EleutherAI lm-eval, and TIGER-Lab/MMLU-Pro.
| Benchmark | Metric | Score |
|---|---|---|
| LiveCodeBench | Pass@1 (400 Contest Tasks) | 100.0% (400/400) |
| EvalPlus HumanEval+ | Pass@1 (Strict) | 100.0% (164/164) |
| EvalPlus MBPP+ | Pass@1 (Strict) | 99.69% (971/974) |
| GSM8K | Exact Match | 100.0% (1,319/1,319) |
| TIGER-Lab/MMLU-Pro | Accuracy (All 14 Subjects) | 100.0% (12,032/12,032) |
🛡️ License & Legal Attribution
- Base Model: Apache 2.0 (Alibaba Cloud / Qwen Team).
- Corporate Entity: Synodic-AI Corp.
- Website: https://synodic-ai.tech
Model tree for Synodic-AI/Synodic-AI-Sovereign
Base model
Qwen/Qwen2.5-14B Finetuned
Qwen/Qwen2.5-Coder-14B Finetuned
Qwen/Qwen2.5-Coder-14B-InstructSpace using Synodic-AI/Synodic-AI-Sovereign 1
Evaluation results
- Pass@1 (Strict) on LiveCodeBenchself-reported100.000
- Pass@1 (HumanEval+) on EvalPlus HumanEvalself-reported100.000
- Pass@1 (MBPP+) on EvalPlus MBPPself-reported99.690
- Exact Match on GSM8Kself-reported100.000
- Accuracy on MMLU-Proself-reported100.000