Comment

Comment by barrell

Meta Llama 3 8B vs Claude Haiku according to their press releases if anyone else was curious

| Benchmark | Llama3 8B | Claude Haiku |

| ------------- | ----------- | ------------ |

| MMLU ____ | 68.4 ____ | 75.2 _______ |

| GPQA ____ | 34.2 ____ | 33.3 _______ |

| HumanEval | 62.2 ____ | 75.9 _______ |

| GSM-8K __ | 79.6 ____ | 88.9 _______ |

| MATH ____ | 30.0 ____ | 40.9 _______ |

parent barrell 2 years ago

Discussion

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton