Llama 3.3 70B

Meta latest generation open-weights model with 128K context. 70B parameters. Improved instruction following (90/100) and reasoning (87/100). Best Llama 70B variant.

Metrics

Performance

ECI Score127.285

Epoch Capabilities Index score

ECI Normalized78

Normalized ECI score (0-100)

Context Window128,000

Maximum input token capacity

Training Compute6.9e24 FLOP

Total compute used during training

Info

8

Score Breakdown

code

20%77

Code generation, understanding, and debugging

math

20%75

Mathematical reasoning and problem solving

reasoning

15%76

Multi-step logical reasoning

intelligence

30%78

Overall reasoning and task completion ability

instruction following

79

Compatibility(8)

Sources(3)

llama.meta.comhuggingface.coepoch.ai

Developed by

Meta AI