Text-to-Image Foundation Models (Diffusion and Flow Transformers)

Large models that generate and edit images from text prompts.

Tier Grades

Tier A: Frontier AI labs

feasible

Frontier and big-tech labs train these models with large teams, for example 39 authors on Qwen-Image.

Tier B: Indonesian AI labs

not feasible

Pretraining a competitive model is not feasible for an Indonesian lab.

Tier C: Institution units such as ITB

stretch

A university unit can fine-tune open weights, for example with LoRA.

Tier D: Organisations such as SEMAI (PoT, PoC, PoV)

feasible

Released models and APIs support marketing and design proofs of concept.

Where this comes from

Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value. Each rationale cites the anchors below.

Anchors

ComputeTraining compute is not in the sources read.
DataNot in the sources read; Qwen-Image trains on text-to-image, text-image-to-image and image-to-image reconstruction tasks.
Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 39 authors on the Qwen-Image report (arXiv 2508.02324).
Time horizonMonths: leaderboards change monthly.
Shelf-life signalsA new open-weight model tops the arena, Licensing and copyright rulings restrict training data
Pipeline roleTier A trains and releases models; Tier C fine-tunes for local styles; Tier D uses APIs and open weights in content pilots.

Where this comes from

Team size is a proxy (author counts of exemplar papers), not a headcount.

Who and What

Exemplars

Qwen-Image Black Forest Labs

FLUX.1 [dev] Stability AI

Stable Diffusion 3.5 Google

Nano Banana Pro (Gemini 3 Pro Image)

Observed actors

Alibaba (Qwen) Black Forest Labs

Black Forest Labs Stability AI

Stability AI Google

Google

Linked models

Nano Banana Pro (Gemini 3 Pro Image)image generationexemplar

Qwen-Imageimage generationexemplar

FLUX.1 [dev]image generationexemplar

Stable Diffusion 3.5image generationexemplar

Nano Banana 2.1image generationproduced by

Qwen-Image-2.1image generationproduced by

GPT Image 2.5 Flareimage generationproduced by

GPT Image 2.5 Sunburstimage generationproduced by

MAI-Image-2.6image generationproduced by

Grok Imagine Image 2.0image generationproduced by

Midjourney V8.2image generationproduced by

FLUX 3 Imageimage generationproduced by

Qwen-Image-3.0image generationproduced by

Qwen-Image-3.0-Proimage generationproduced by

Seedream 5.0 Proimage generationproduced by

Muse Imageimage generationproduced by

Krea 2 Rawimage generationproduced by

Krea 2 Turboimage generationproduced by

Ideogram 4.0image generationproduced by

Nano Banana 2 Liteimage generationproduced by

AsymFLUX.2 kleinimage generationproduced by

Krea 2 Largeimage generationproduced by

Midjourney V8.1image generationproduced by

GPT Image 2image generationproduced by

+107

Midjourney V8.0 alphaimage generationproduced by

Riverflow 2.0image generationproduced by

Nano Banana 2 (Gemini 3.1 Flash Image)image generationproduced by

Niji 7image generationproduced by

GLM-Imageimage generationproduced by

Qwen-Image-2512image generationproduced by

Qwen-Image-Layeredimage generationproduced by

Seedream 4.5image generationproduced by

GPT Image 1.5image generationproduced by

P-Image (P-Image 1.0)image generationproduced by

Ovis-Image 7Bimage generationproduced by

Z-Image Turboimage generationproduced by

FLUX.2image generationproduced by

FLUX.2 [dev]image generationproduced by

FLUX.2 [klein] 4Bimage generationproduced by

FLUX.2 [klein] 9Bimage generationproduced by

FLUX.2 [pro]image generationproduced by

FLUX.2 [flex]image generationproduced by

Bria FIBOimage generationproduced by

Firefly Image 5image generationproduced by

HunyuanImage 3.0image generationproduced by

HunyuanImage 3.0 Instructimage generationproduced by

HunyuanImage 2.1image generationproduced by

Seedream 4.0image generationproduced by Chroma1-HDimage generationproduced by

Qwen-Imageimage generationproduced by

FLUX.1 Krea Devimage generationproduced by

STARFlowimage generationproduced by

HiDream-I1image generationproduced by

FLUX.1 Kontextimage generationproduced by

Seedream 3.0image generationproduced by

HiDream-I1-Devimage generationproduced by

Firefly Image 4image generationproduced by

Midjourney V7image generationproduced by

MiniMax Image-01image generationproduced by

Luma Photonimage generationproduced by

Auroraimage generationproduced by

SANAimage generationproduced by

SANA 1.6Bimage generationproduced by

SANA 0.6Bimage generationproduced by

FLUX1.1 [pro]image generationproduced by

Ideogram v2image generationproduced by

FLUX.1image generationproduced by

FLUX.1 [schnell]image generationproduced by

Imagen 3image generationproduced by

Hunyuan-DiTimage generationproduced by Juggernaut-XL v9image generationproduced by

Playground v2.5image generationproduced by

Recraft 20Bimage generationproduced by

Kolorsimage generationproduced by

Midjourney V6image generationproduced by

DeciDiffusion 1.0image generationproduced by

DALL-E 3image generationproduced by

Kandinsky 2.2image generationproduced by

LDM3D3Dproduced by Openjourneyimage generationproduced by

DALL-E 2image generationproduced by

GLIDEimage generationproduced by

Firefly Imageimage generationproduced by

Z-Image Baseimage generationproduced by

FLUXimage generationproduced by

FLUX 2 Maximage generationproduced by

Seedreamimage generationproduced by

Seedream 5.0 Liteimage generationproduced by

Imagenimage generationproduced by

Imagen 4image generationproduced by

Gemini Nano Banana 2.1image generationproduced by

Nano Bananaimage generationproduced by

HiDreamimage generationproduced by

HiDream-O1-Imageimage generationproduced by

HiDream-O1-Image-1.5image generationproduced by

Ideogramimage generationproduced by

Ideogram 3image generationproduced by

Krea 2image generationproduced by

Uni-1image generationproduced by

Uni-1.1image generationproduced by

Uni-1.1-maximage generationproduced by

MAI-Imageimage generationproduced by

MAI-Image-2.5image generationproduced by

Midjourneyimage generationproduced by

Midjourney V6.1image generationproduced by

GPT Imageimage generationproduced by

Recraftimage generationproduced by

Recraft V2image generationproduced by

Recraft V3image generationproduced by

Recraft V4image generationproduced by

Recraft V4.1image generationproduced by

Reve 2.0image generationproduced by

Reve 2.1image generationproduced by

Grok Imagineimage generationproduced by

Grok Imagine Imageimage generationproduced by

SDXLimage generationproduced by

Stable Diffusionimage generationproduced by

HunyuanImageimage generationproduced by

HunyuanImage 3.5 Previewimage generationproduced by

Stable Image Ultraimage generationproduced by

Stable Image Coreimage generationproduced by

Bria image modelsimage generationproduced by

Amazon Nova Canvasimage generationproduced by

DALL-Eimage generationproduced by

Bria 3.2image generationproduced by

GPT Image 1image generationproduced by

GPT Image 1 miniimage generationproduced by

Ideogram 4.5image generationproduced by

Leonardo Phoenix 1.0image generationproduced by

Leonardo Lucid Originimage generationproduced by

Arrow 1.1image generationproduced by

Papers

High-Resolution Image Synthesis with Latent Diffusion ModelsRobin Rombach et al., CVPR 2022 (per arXiv comment), 2021 founding
FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent SpaceBlack Forest Labs et al., arXiv preprint, 2025 state of the art
Scaling Rectified Flow Transformers for High-Resolution Image SynthesisPatrick Esser et al., arXiv preprint, 2024 state of the art
SDXL: Improving Latent Diffusion Models for High-Resolution Image SynthesisDustin Podell et al., arXiv preprint, 2023 state of the art
Hierarchical Text-Conditional Image Generation with CLIP LatentsAditya Ramesh et al., arXiv preprint, 2022 state of the art

Where this comes from

The papers to read first: where the direction began, its current best, and a survey when one exists.

Sources

arxiv.org/abs/2508.02324 https://arxiv.org/abs/2508.02324

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Text-to-Image Foundation Models (Diffusion and Flow Transformers)", https://aurion.semai.id/topics/text-to-image-foundation-models/, records of October 2026.

Text-to-Image Foundation Models (Diffusion and Flow Transformers)

Tier Grades

Tier A: Frontier AI labs

Tier B: Indonesian AI labs

Tier C: Institution units such as ITB

Tier D: Organisations such as SEMAI (PoT, PoC, PoV)

Anchors

Who and What

Links

Models 131

Papers

Sources