Comprehensive performance analysis of SnowflakeCore-G1-Tiny models across multiple benchmarks
| Metric | Base Model | Instruct Model |
|---|
Base model generates text 7% faster than the instruction-tuned variant
Instruction model produces slightly shorter responses on average
Base model shows higher vocabulary diversity (16.6% vs 12.9%)
| Sequence Length | Base Model (GB) | Instruct Model (GB) |
|---|
Both models show identical memory usage patterns across different sequence lengths
Data extracted from SnowflakeCore model benchmarks • All metrics verified through synthetic testing