Back to models

Llama 3.3 70B

ECI#169
126
Meta AI
Dec. 6, 2024
Open weights

Llama 3.3 70B is an open-weights model from Meta AI, released December 6, 2024. It has an ECI score of 126 and ranks 169th of 239 tracked models. At the time of its release, it ranked 13th of 83 tracked models.

Llama 3.3 70B vs select alternatives

DomainBenchmark
Llama 3.3 70B
Meta AI
Dec. 6, 2024
Open weights
Llama 3.1-70B
Meta AI
Jul. 23, 2024
Open weights
Llama 3.1-405B
Meta AI
Jul. 23, 2024
Open weights
Claude 3.5 Haiku
Anthropic
Oct. 22, 2024
Closed weights
ECI scoreBest score 174
126125128125
Ranking#169#175#161#173
Conceptual reasoning
SimpleBenchBest score 82%
20%
n/a
23%
n/a
Long context
ForecastBenchBest score 66.7
61.0—62.3—
Mathematics
MATH Level 5Best score 98%
42%
37%
50%
46%
OTIS Mock AIME 2024-2025Best score 100%
5%
4%
10%
4%
Science
CritPt v1Best score 36%
0%
n/a
0%
0%
SciCodeBest score 67%
26%
n/a
n/a
27%
GPQA DiamondBest score 96%
47%
44%
51%
38%
Software engineering
WeirdML (v2)Best score 94%
14%
9%
21%
31%
Other
LiveBenchBest score 82%
50%
n/a
n/a
43%

Model metadata

General
OrganizationMeta AI
Release dateDec. 6, 2024
Model accessibilityOpen weights
CountryUnited States of America
Knowledge cutoff2023-12-01
Provider notesMeta Llama 3.3 multilingual large language model (LLM)
Pricing
Input price$0.442 / 1M tokens
Output price$0.71 / 1M tokens
Training
Parameters7.0e10
Training compute6.9e24 FLOP
Training hardwareNVIDIA H100 SXM5 80GB
Training timeUnknown
Training dataset size1.5e13 datapoints