Skip to content
Updated minutes ago
OpenAI

DALL-E 3

OpenAI · dense · 3.5B parameters · 4,096 context

Quality
50.0

Parameters

3.5B

Context Window

4K tokens

Architecture

Dense

Best GPU

RTX 4070 Ti

Cheapest API

$40.00/M

Intelligence Brief

DALL-E 3 is a 3.5B parameter DENSE model from OpenAI, featuring Multi-Head Attention (MHA) with 32 layers and 2,560 hidden dimensions. With a 4,096 token context window, it supports vision, multilingual. The most cost-effective API deployment is via openai at $40.00/M output tokens. For self-hosted inference, RTX 4070 Ti delivers optimal throughput at $237/month.

Architecture Details

TypeDENSE
Total Parameters3.5B
Active Parameters3.5B
Layers32
Hidden Dimension2,560
Attention Heads32
KV Heads32
Head Dimension80
Vocab Size100,000

Memory Requirements

BF16 Weights

7.0 GB

FP8 Weights

3.5 GB

INT4 Weights

1.8 GB

KV-Cache per Token163840 bytes
Activation Estimate1.00 GB

GPU Compatibility Matrix

DALL-E 3 is compatible with 100% of GPU configurations across 41 GPUs at 3 precision levels.

BF16 (Full)
FP8 (Half)
INT4 (Quarter)
Blackwell(7 GPUs)
B200 NVL (pair)360GB
B300288GB
B100 SXM192GB
GB200 NVL72 (per GPU)192GB
Hopper(7 GPUs)
H100 NVL 94GB (per GPU pair)188GB
H200 SXM141GB
H2096GB
GH20096GB
Ada Lovelace(11 GPUs)
L40S48GB
L4048GB
RTX 6000 Ada48GB
L2048GB
Ampere(16 GPUs)
A100 80GB SXM80GB
A100 80GB PCIe80GB
A1664GB
RTX A600048GB
Legend:No fitVery tightTightModerateGoodExcellent

GPU Recommendations

RTX 4070 Tioptimal

BF16 · 1 GPU · vllm

100/100

score

Throughput

349.9 tok/s

Latency (ITL)

2.9ms

Est. TTFT

0ms

Cost/Month

$237

Cost/M Tokens

$0.26

Use this config →
RTX 3080optimal

BF16 · 1 GPU · vllm

100/100

score

Throughput

527.6 tok/s

Latency (ITL)

1.9ms

Est. TTFT

0ms

Cost/Month

$133

Cost/M Tokens

$0.10

Use this config →
RTX 4070 Superoptimal

BF16 · 1 GPU · vllm

100/100

score

Throughput

349.9 tok/s

Latency (ITL)

2.9ms

Est. TTFT

0ms

Cost/Month

$209

Cost/M Tokens

$0.23

Use this config →

Deployment Options

API

API Deployment

openai

$40.00/M

output tokens

Self-Hosted

Single GPU

RTX 4070 Ti

$237/mo

Min VRAM: 4 GB

Scale

Multi-GPU

RTX 3070 x2

438.8 tok/s

TP· $171/mo

API Pricing Comparison

ProviderInput $/MOutput $/MBadges
openai$40.00$40.00
Cheapest

Cost Analysis

ProviderInput $/MOutput $/M~Monthly Cost
openaiBest Value$40.00$40.00$400

Cost per 1,000 Requests

Short (500 tok)

$28.00

via openai

Medium (2K tok)

$112.00

via openai

Long (8K tok)

$400.00

via openai

Performance Estimates

Throughput by GPU

RTX 4070 Ti
349.9 tok/s
RTX 3080
527.6 tok/s
RTX 4070 Super
349.9 tok/s

VRAM Breakdown (RTX 4070 Ti, BF16)

Weights
KV
Act
Weights 7.0 GBKV-Cache 5.4 GBActivations 8.0 GBOverhead 0.6 GB

Capabilities

Features

Tool Use Vision Code Math Reasoning Multilingual Structured Output

Supported Frameworks

Supported Precisions

BF16 (default)

Where to Deploy DALL-E 3

Similar Models

Llama 3.2 3B

3.21B params · dense

Quality: 55

from $0.06/M

Larger context, CheaperCompare →
Larger context, Higher qualityCompare →
Larger context, Higher qualityCompare →

Frequently Asked Questions

How much VRAM does DALL-E 3 need for inference?

DALL-E 3 requires approximately 7.0 GB of VRAM at BF16 precision, 3.5 GB at FP8, or 1.8 GB at INT4 quantization. Additional VRAM is needed for KV-cache (163840 bytes per token) and activations (~1.00 GB).

What is the best GPU for DALL-E 3?

The top recommended GPU for DALL-E 3 is the RTX 4070 Ti using BF16 precision. It achieves approximately 349.9 tokens/sec at an estimated cost of $237/month ($0.26/M tokens). Score: 100/100.

How much does DALL-E 3 inference cost?

DALL-E 3 API inference starts from $40.00/M input tokens and $40.00/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.