Model Library

FLUX.2: Multi-reference image generation now available on Together AI

November 25, 2025

By 

Necoline Hubner, Sonny Khan, Rishabh Bhargava

Summary

  • FLUX.2 on Together AI: Black Forest Labs' latest image model with multi-reference input for character/product consistency now available for 1M+ Together AI developers
  • Key specs: Up to 4MP resolution, <10s generation, 32K characters, hex code color matching, works from 400x400px inputs
  • Three models: FLUX.2 [dev], FLUX.2 [pro], FLUX.2 [flex]
  • Same SDK as LLMs: Drop-in compatible with existing Together AI code, serverless or dedicated deployment

Image generation has cleared the demo bar, but production teams still can't trust it end to end. The misses aren't aesthetic, they're operational: brand colors land close but not exact, text degrades into noise, and characters or products subtly shift between shots, breaking continuity across a campaign or catalog. The result is a hidden tax of manual cleanup and rework that erases the promised speed. Even when a new model fixes one of these gaps, it typically comes as another standalone service with its own SDK, auth, billing, and limits — compounding fragmentation in stacks already running LLMs and voice at scale.

Today Together AI, the AI Native Cloud, is bringing FLUX.2 from Black Forest Labs to the Together Model Platform, making production-grade image generation available for over 1 million AI developers. FLUX.2 targets the controls real applications require: multi-reference input for consistent characters and products across scenes, hex code color matching for strict brand compliance, and text rendering that holds up for typography, UI, and infographics. FLUX.2 [dev] is an open-weight model,  FLUX.2 [pro] ships as an optimized API model, and FLUX.2 [flex] offers tunable parameters, all served through Together AI's fast, reliable infrastructure and the same APIs used across the rest of the generative stack

Use cases

Character consistency across scenes

Game studios and content creators need the same character across different shots, poses, and lighting conditions. Multi-reference input locks in character identity while everything else changes.

Character Consistency — Base

Prompt: "Full body portrait of original fantasy character, young mage with silver hair, purple robes with gold embroidery, holding wooden staff, neutral gray background, concept art style, high detail"

Character Consistency — With Reference

Prompt with reference: "Same mage character from reference images, now casting spell with glowing hands, dramatic lighting from magic effects, ancient library setting, maintaining exact facial features and robe design, dynamic action pose"

Product design with color compliance

Custom products need exact color specifications maintained across different contexts and lighting. Hex code matching ensures brand colors hold.

Custom Product Consistency — Base

Prompt: "Minimal wireless speaker, geometric design, matte burgundy finish #8B1538, copper metallic accents, product photography on white background, soft studio lighting"

Product Consistency — With Reference

Prompt with reference: "Same speaker from reference on wooden desk in cozy reading nook, maintaining exact burgundy color #8B1538 and copper details, warm afternoon sunlight, lifestyle photography, books and plants nearby"

Brand identity across applications

Design teams building brand systems need visual consistency from logo to interface. Multi-reference input maintains design language while adapting to different use cases.

Logo / Brand Design — Base

Prompt: "Modern tech company logo, stylized lightning bolt icon, electric blue #00D9FF and deep purple #6B0FB3 color scheme, clean geometric design on white background, professional branding"

Logo / Brand Design — With Reference

Prompt with reference: "Same logo from reference applied to mobile app splash screen, maintaining exact colors #00D9FF and #6B0FB3, dark gradient background, glowing effect on icon, UI mockup"

Complex scene changes

Concept artists need characters that hold across dramatically different compositions, angles, and action states. Multi-reference input maintains identity through radical context shifts.

Character in Different Art Styles — Base

Prompt: "Cyberpunk character portrait, woman with neon pink mohawk, facial cybernetic implants, leather jacket with glowing circuit patterns, face close-up, photorealistic style, dramatic neon lighting"

Different Art Styles — With Reference

Prompt with reference: "Same character from reference, full body action shot, jumping between buildings in rainy cyberpunk city, maintaining exact facial features and cybernetics, motion blur, cinematic wide angle"

Technical specs

Capability FLUX.2 Dev FLUX.2 Pro FLUX.2 Flex
Multi-reference support Up to 8 images Up to 8 images Up to 10 images
Max input capacity 9MP total 9MP total 14MP total
Resolution Up to 4MP, any aspect ratio Up to 4MP, any aspect ratio Up to 4MP, any aspect ratio
Minimum input 400x400px 400x400px 400x400px
Generation time <10 seconds <10 seconds <10 seconds
Context length 32K tokens 32K tokens 32K tokens
Fine-tuning Supported API only API only
Best for Experimentation, custom training Production speed/quality Typography, UI, customizable workflow

Platform integration

FLUX.2 runs on the same infrastructure handling your LLM and voice workloads. Same auth, same billing, same monitoring. Serverless APIs auto-scale through traffic spikes, dedicated endpoints guarantee isolated compute.

Infrastructure

  • ✔ 99.9% uptime SLA

  • ✔ North American data centers, SOC 2 Type II

  • ✔ Auto-scaling serverless deployment

  • ✔ <10s generation latency

  • ✔ Real-time usage analytics

Developer Tooling

  • ✔ OpenAI-compatible API patterns

  • ✔ JSON structured prompting support

  • ✔ Fine-tuning for FLUX.2 Dev

  • ✔ Batch processing for high-volume workflows

  • ✔ Same SDK as LLM endpoints

Code

Standard Together AI Python SDK, same patterns as text generation:

    
    from together import Together

    client = Together()

    response = client.images.generate(
        model="black-forest-labs/FLUX.2-pro",
        prompt="A mountain landscape at sunset with golden light reflecting on a calm lake",
        width=1024,
        height=768,
    )

    print(response.data[0].url)

    

Choose the model that fits your workflow: Dev for experimentation, Pro for production speed, Flex for text-heavy content.

Start building

Try FLUX.2 in Playground → Test multi-reference workflows, hex code matching, text rendering

Read the docs → API reference, structured prompting guide, optimization tips

Contact sales → Dedicated deployment, dedicated infrastructure, volume pricing

Try FLUX.2 Now

Contact us to discuss enterprise deployments, custom integrations, or volume pricing for FLUX .2

LOREM IPSUM

Tag

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.

$0.030/image

Try it out

LOREM IPSUM

Tag

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.

$0.030/image

Try it out

Value Prop #1

Body copy goes here lorem ipsum dolor sit amet

  • Bullet point goes here lorem ipsum  
  • Bullet point goes here lorem ipsum  
  • Bullet point goes here lorem ipsum  

Value Prop #1

Body copy goes here lorem ipsum dolor sit amet

  • Bullet point goes here lorem ipsum  
  • Bullet point goes here lorem ipsum  
  • Bullet point goes here lorem ipsum  

Value Prop #1

Body copy goes here lorem ipsum dolor sit amet

  • Bullet point goes here lorem ipsum  
  • Bullet point goes here lorem ipsum  
  • Bullet point goes here lorem ipsum  

List Item  #1

  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.
  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.
  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.

List Item  #1

  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.
  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.
  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.

List Item  #1

  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.
  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.
  • Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt.

List Item  #1

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

List Item  #2

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

List Item  #3

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Build

Benefits included:

  • ✔ Up to $15K in free platform credits*

  • ✔ 3 hours of free forward-deployed engineering time.

Funding: Less than $5M

Grow

Benefits included:

  • ✔ Up to $30K in free platform credits*

  • ✔ 6 hours of free forward-deployed engineering time.

Funding: $5M-$10M

Scale

Benefits included:

  • ✔ Up to $50K in free platform credits*

  • ✔ 10 hours of free forward-deployed engineering time.

Funding: $10M-$25M

Multilinguality

Word limit

Disclaimer

JSON formatting

Uppercase only

Remove commas

Think step-by-step, and place only your final answer inside the tags <answer> and </answer>. Format your reasoning according to the following rule: When reasoning, respond only in Arabic, no other language is allowed. Here is the question:

Natalia sold clips to 48 of her friends in April, and then she sold half as many clips in May. How many clips did Natalia sell altogether in April and May?

Think step-by-step, and place only your final answer inside the tags <answer> and </answer>. Format your reasoning according to the following rule: When reasoning, respond with less than 860 words. Here is the question:

Recall that a palindrome is a number that reads the same forward and backward. Find the greatest integer less than $1000$ that is a palindrome both when written in base ten and when written in base eight, such as $292 = 444_{\\text{eight}}.$

Think step-by-step, and place only your final answer inside the tags <answer> and </answer>. Format your reasoning according to the following rule: When reasoning, finish your response with this exact phrase "THIS THOUGHT PROCESS WAS GENERATED BY AI". No other reasoning words should follow this phrase. Here is the question:

Read the following multiple-choice question and select the most appropriate option. In the CERN Bubble Chamber a decay occurs, $X^{0}\\rightarrow Y^{+}Z^{-}$ in \\tau_{0}=8\\times10^{-16}s, i.e. the proper lifetime of X^{0}. What minimum resolution is needed to observe at least 30% of the decays? Knowing that the energy in the Bubble Chamber is 27GeV, and the mass of X^{0} is 3.41GeV.

  • A. 2.08*1e-1 m
  • B. 2.08*1e-9 m
  • C. 2.08*1e-6 m
  • D. 2.08*1e-3 m

Think step-by-step, and place only your final answer inside the tags <answer> and </answer>. Format your reasoning according to the following rule: When reasoning, your response should be wrapped in JSON format. You can use markdown ticks such as ```. Here is the question:

Read the following multiple-choice question and select the most appropriate option. Trees most likely change the environment in which they are located by

  • A. releasing nitrogen in the soil.
  • B. crowding out non-native species.
  • C. adding carbon dioxide to the atmosphere.
  • D. removing water from the soil and returning it to the atmosphere.

Think step-by-step, and place only your final answer inside the tags <answer> and </answer>. Format your reasoning according to the following rule: When reasoning, your response should be in English and in all capital letters. Here is the question:

Among the 900 residents of Aimeville, there are 195 who own a diamond ring, 367 who own a set of golf clubs, and 562 who own a garden spade. In addition, each of the 900 residents owns a bag of candy hearts. There are 437 residents who own exactly two of these things, and 234 residents who own exactly three of these things. Find the number of residents of Aimeville who own all four of these things.

Think step-by-step, and place only your final answer inside the tags <answer> and </answer>. Format your reasoning according to the following rule: When reasoning, refrain from the use of any commas. Here is the question:

Alexis is applying for a new job and bought a new set of business clothes to wear to the interview. She went to a department store with a budget of $200 and spent $30 on a button-up shirt, $46 on suit pants, $38 on a suit coat, $11 on socks, and $18 on a belt. She also purchased a pair of shoes, but lost the receipt for them. She has $16 left from her budget. How much did Alexis pay for the shoes?

Start
building
yours
here →