Llama 3.3
Text-only 70B instruction-tuned model with enhanced performance for text-based applications.
Llama 3.3 is Meta's text-only 70B instruction-tuned multilingual large language model release, optimized for multilingual dialogue and text-based applications.
Pricing
Model profile · Current snapshot
Input / 1M
$0.655
Output / 1M
$0.72
Blended / 1M
$0.671
Output TPS
83.67
TTFT (s)
0.64
Model Intelligence
Context window
128,000 tokens
Arena ranking
9
Benchmarkable
Yes
Model level
release
Intelligence Index
9.3
Coding Index
11.9
Math Index
7.7
MMLU Pro
0.71
GPQA
0.5
HLE
0.04
LiveCodeBench
0.29
SciCode
0.26
MATH-500
0.77
AIME
0.3
AIME 2025
0.08
IFBench
0.47
LCR
0.16
TerminalBench Hard
0.03
TAU2
0.27
Recent stories
0 linked stories
No linked stories yet.