Curiosity · AI Model

Recraft V3

Recraft V3 (codename 'red_panda') is the third-generation image model from London-based Recraft AI, released in October 2024. It briefly topped the Artificial Analysis text-to-image leaderboard and is widely regarded as best-in-class for accurate in-image text rendering, long-caption alignment, and vector-style illustration. Recraft exposes V3 through its own design tool and a REST API, making it popular with brand and marketing teams that need reliable typography, logos, and icon sets.

Model specs

Vendor: Recraft AI
Family: Recraft
Released: 2024-10
Context window: 1 tokens
Modalities: text, vision
Input price: n/a
Output price: n/a
Pricing as of: 2026-04-20

Strengths

Industry-leading in-image text rendering
Unique vector / illustration mode
Strong long-caption prompt alignment
Integrated design-tool UX

Limitations

Closed weights — API only, credit-based pricing
Weaker on photoreal portraits vs Flux / Imagen 3
No fine-tuning exposed publicly
Smaller community and fewer tutorials than SDXL

Use cases

Marketing and brand visuals with accurate typography
Icon, logo, and illustration generation
Vector-style creative work
Ad creative and social imagery

Benchmarks

Benchmark	Score	As of
Artificial Analysis Image Arena	#1 at launch (Oct 2024)	2024-10
Text-in-image fidelity (human eval)	industry-leading	2024-10

Frequently asked questions

What is Recraft V3?

Recraft V3 is Recraft AI's flagship text-to-image model released in October 2024, known especially for accurate in-image text rendering and vector-style illustrations.

Is Recraft V3 open source?

No — it is closed-weights and accessed via Recraft's web app or REST API.

What is Recraft V3 best at?

Text-in-image fidelity, brand/marketing visuals with legible typography, and vector/illustration styles — areas where most other models still struggle.

Sources

Recraft V3 launch blog — accessed 2026-04-20
Artificial Analysis image leaderboard — accessed 2026-04-20