Back to models

Falcon-40B

ECI#227
104
Technology Innovation Institute
May 25, 2023
Open weights

Falcon-40B is an open-weights model from Technology Innovation Institute, released May 25, 2023. It has an ECI score of 104 and ranks 227th of 249 tracked models. At the time of its release, it ranked 4th of 10 tracked models.

Input price—
Output price—
Parameters4.0e10
Training compute2.4e23 FLOP

Falcon-40B vs select alternatives

Reasoning
DomainBenchmark
Falcon-40B
Technology Innovation Institute
May 25, 2023
Open weights
Falcon-7B
Technology Innovation Institute
Apr. 24, 2023
Open weights
GPT-4 (Mar 2023)
OpenAI
Mar. 14, 2023
Closed weights
LLaMA-65B
Meta AI
Feb. 24, 2023
Open weights
ECI scoreBest score 166
10495126110
Ranking#227#238#173#216
Conceptual reasoning
BBHBest score 89.2%
37.1%
28.8%
n/a
58.4%
Language understanding
LAMBADABest score 87.2%
77.3%
74.9%
n/a
77.7%
PIQABest score 88.7%
83.0%
80.3%
n/a
82.8%
BoolQBest score 91.2%
83.1%
75.3%
n/a
87.1%
HellaSwagBest score 95.3%
85.3%
78.1%
95.3%
84.2%
Mathematics
GSM8KBest score 94.5%
33.8%
6.8%
92.0%
54.4%
Science
OpenBookQABest score 88.0%
56.6%
51.6%
n/a
60.2%
ARC AI2Best score 95.3%
61.9%
47.9%
n/a
69.5%
World knowledge
MMLUBest score 88.1%
56.9%
35.0%
86.4%
63.4%

Model metadata

General
OrganizationTechnology Innovation Institute
Release dateMay 25, 2023
Model accessibilityOpen weights
CountryUnited Arab Emirates
Knowledge cutoffUnknown
Provider notesAbu Dhabi-based Technology Innovation Institute Introduces Falcon LLM: Foundational Large Language Model (LLM) outperforms GPT-3 with 40 Billion Parameters
Pricing
Input price—
Output price—
Training
Parameters4.0e10
Training compute2.4e23 FLOP
Training hardwareNVIDIA A100
Training time60 days
Training dataset size1.0e12 datapoints