benchmarking/tinystories_5m_eval_metrics.json

{
  "model": "5M",
  "num_samples": 100,
  "temperature": 0.8,
  "top_k": 50,
  "metrics": {
    "ttft_ms": 4.12,
    "tps": 530.8,
    "total_latency_ms": 61.0,
    "tokens_generated": 32.2,
    "perplexity": 4.26,
    "avg_token_prob": 0.4944,
    "repetition_rate": 0.095,
    "coherence_length": 22.4,
    "vocab_diversity": 0.8346
  },
  "evaluated_at": "2026-06-21 09:19:58"
}