Grok Imagine Image 2.0 Debuts at No.4 on Artificial Analysis Image Leaderboard

Artificial Analysis released several evaluations on September 19, announcing that xAI's Grok Imagine Image 2.0 has entered its image leaderboard: ranked 4th overall in text-to-image (Elo 1154, tied with Microsoft MAI-Image-2.6 at positions 4–5), making it the highest-ranked model outside of OpenAI, up 14 places from the previous generation grok-imagine-image-qualit; it also ranks 4th in image editing, while costing only about 30% of GPT Image. Overall, it is a new-generation image model with outstanding quality-per-dollar.

Confirmed

• 4th overall in text-to-image with Elo 1154, tied with Microsoft MAI-Image-2.6 at positions 4–5; up 14 places from the previous generation, the highest ranking outside of OpenAI.
• Broken down across 9 capabilities: knowledge (real landmarks, species, science and common-sense facts) and text rendering (long text, small fonts, symbols, artistic lettering) are closest to the frontier, followed by reasoning, materials, and complex composition-related capabilities.
• Measured across 10 use cases: text-to-image is closest to the top in animation & games and productivity & knowledge work, followed by consumer uses (book covers, stock imagery, editorial illustration) and UI/UX design; it improves over the previous generation across all use cases.
• Image editing: across 7 editing action categories, it is closest to the frontier in scene and style editing (relighting, restyling, background and overall treatment changes), followed by identity-preserving edits; across 10 use-case dimensions, animation & games (game assets, concept art, character design, comics) is strongest, followed by live-action film/TV and knowledge work.
• Independent speed test: generating a 1024x1024 medium-quality image takes a median of 84 seconds, faster than GPT Image 2 (high) at 121 seconds, but slower than the fastest models in the same tier.
• Pricing: $80 per 1,000 2K images, sitting on the Pareto frontier of the quality-price curve, between Microsoft MAI-Image-2.6 and OpenAI GPT Image 2.5 Flare.

Why it matters

• Grok Imagine Image 2.0 enters the first tier at a price far below GPT Image and occupies the quality-price Pareto frontier, offering a new low-cost option for pay-as-you-go developers and content producers.
• Its strengths in knowledge and text rendering mean it has practical value in productivity, knowledge work, and text-heavy design scenarios, rather than being limited to stylized generation.

2026-09-19 ~ 2026-09-19 · 8 related posts

• [source] Grok Imagine Image 2.0 Climbs 14 Spots to #4 on Text-to-Image Leaderboard, Best Non-OpenAI Model — ArtificialAnlys · 2026-09-19
• [source] Grok Imagine 2.0 Priced at $60-80 per 1K Images, on Quality-vs-Price Pareto Frontier — ArtificialAnlys · 2026-09-19
• Grok Imagine 2.0 Takes a Median 84s Per 1024x1024 Image, Slower Than Rivals at Its Quality Tier — ArtificialAnlys · 2026-09-19
• Grok Imagine 2.0 Strongest at Knowledge, Text Rendering and Reasoning in Text-to-Image — ArtificialAnlys · 2026-09-19
• Grok Imagine 2.0 Closes the Gap to the Frontier Across All 10 Text-to-Image Use Cases — ArtificialAnlys · 2026-09-19
• Where Grok Imagine 2.0 Edits Best: Scene & Style, Identity-Preserving, Restoration — ArtificialAnlys · 2026-09-19
• [source] Grok Imagine Image 2.0 debuts at #4 on image leaderboard, 1/3 the price of GPT Image — ArtificialAnlys · 2026-09-19

1 near-duplicate retellings: ArtificialAnlys