Fusion in Encoder

Closed weights Samsung 330M parameters November 2022

No estimate

No hardware requirements for this model

The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.

On record

Full specification

Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.

Origin

Who built this model, where, and when it was published.

Organisation
Samsung
Organisation type
Industry
Country
Korea (Republic of)
Published
18 November 2022
Authors
Akhil Kedia, Mohd Abbas Zaidi, Haejun Lee

What it does

The problem areas the model was built for. A model can carry several of each.

Domain
Language
Task
Question answering, Language modeling/generation

Size

How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.

Parameters
330M

330M

Training data
960,000 tokens

79k per table 11 (probably number of question-answer pairs)

Training compute

The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.

Training compute
1.3 × 10²⁰ FLOP

"The experiments were run on 8x80GB Nvidia A100s with 800GB RAM and 4x32-core CPUs, and each experiment took around 1 day for NQ and 2 days for TriviaQA with large models. Inference was run on the same system, and took 2 minutes." 2 days * 24 * 3600 * 8 * 312 teraflop/s * 0.3 utilization = 1.3e20

How it was established
Hardware

The training run

What it physically took to train: which chips, how many, for how long, and what that drew from the wall.

Training hardware
NVIDIA A100 SXM4 80 GB
Wall-clock time
48 hours

2 days

Compute cost
$233

Availability

Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.

Weights
Closed — provider access only
Model access
Unreleased
Training code
Unreleased

"The source code is based on the original implementation of FiD (Izacard and Grave, 2021b), which can be found at their Github. The modeling for the fused Electra model was implemented using HuggingFace (Wolf et al., 2020) by modifying the ElectraModel.

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Why it is tracked
SOTA improvement

"Using our proposed method, we outperform the current state-of-the-art method by 2.5 Exact Match score on the Natural Question dataset while using only 25% of parameters and 35% of the latency during inference, and 4.4 Exact Match on WebQuestions dataset"

Record confidence
Likely
Citations
12

Sources

Where this record came from and when it was last checked.

Reference
FiE: Building a Global Probability Space by Leveraging Early Fusion in Encoder for Open-Domain Question Answering
Last updated
25 May 2026

What the numbers mean

About this model

Fusion in Encoder was published by Samsung, in the country recorded as Korea (Republic of), during November 2022. The category the publisher falls under is industry.

It works in the domain of Language, and is recorded as performing the task of question answering, Language modeling/generation.

Its weights were never published, so it can only be reached through its provider. No graphics card changes that.

What went into building it

Producing it required arithmetic totalling around 1.3 × 10²⁰ FLOP, on hardware recorded as NVIDIA A100 SXM4 80 GB. That figure measures what producing the model cost, and has no bearing on how fast it answers.

Training consumed a corpus of around 960,000 tokens of text.

It is tracked in the underlying dataset for one reason in particular: sOTA improvement.

Answers

Fusion in Encoder — common questions

01

Fusion in Encoder— how much compute was used to train it?

Training consumed around 1.3 × 10²⁰ FLOP, on hardware recorded as NVIDIA A100 SXM4 80 GB. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.

02

Fusion in Encoder— what GPU do I need to run it?

None. This is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.

03

Fusion in Encoder— is it open source?

No. Its weights have not been published, so it exists only as a service controlled by its owner.

04

Fusion in Encoder— how many parameters does it have?

It has a parameter count of 330M. 330M. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.

05

Fusion in Encoder— who created it?

It was published by Samsung, based in Korea (Republic of), an organisation categorised as industry.

06

Fusion in Encoder— when was it released?

It was published in November 2022. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

07

Fusion in Encoder— what is it used for?

It works in the domain of Language, and is recorded as handling the task of question answering, Language modeling/generation. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Source

Original publication

Record last updated 25 May 2026

The other direction

Looking at it from the other side?

This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.