0

Tülu 3 70B

Open weights

Allen AI's fully open-source post-trained Llama 3.1 70B with a novel RLVR (RL with verifiable rewards) recipe - data, code, and weights all open.

Family
Tülu 3
Params
70B
Context
8K
Input
Output
Released
Nov 2024

Cite

Notes

Only stored in your browser.

Attribution

Benchmark scores
Vault
Attribution policy →