Can I trust these benchmark results?

We ran the exact same prompt 3-5 times on each model with temperature 0.7 where configurable, then averaged. Scores are editorial — honest, careful, reproducible — but not an academic study with 1,000 runs. The exact prompt is published above so you can verify yourself.

How often is this benchmark updated?

Last run: 2026-09-25. We re-run when a tested model gets a major version bump (Midjourney v6 → v7, Claude 3.5 → 4, etc.) or quarterly if nothing changes. If a new model enters the category (e.g. a new image model competing with Midjourney), it gets added to the leaderboard within 30 days of launch.

Which model is the best value for money on this task?

Stable Diffusion 3.5 + anime LoRA (free local run). For most users, the price-performance winner matters more than the overall winner — if you're running thousands of generations a month, a 10× cheaper model that scores 1-2 points lower is often the right call.

How can I reproduce this benchmark?

The exact prompt is in the "The prompt" section above. Copy it, run it on these 5 models (Midjourney v6.1 niji, NovelAI v3, Stable Diffusion 3.5 + anime LoRA, FLUX 1.1 Pro, DALL-E 3) with temperature 0.7, and score each output on Anime fidelity, Composition, Detail richness, Prompt adherence, Style consistency. Your results will be within one point of ours on most dimensions.

Best AI model for anime illustration

Q: Why did Midjourney v6.1 (--niji mode) win the Image — Anime benchmark?

Midjourney v6.1 (--niji mode) scored highest across the 5 rated dimensions (Anime fidelity, Composition, Detail richness, Prompt adherence, Style consistency). For anime specifically, Midjourney's --niji mode is the clear winner — it was trained with anime-specific weighting. See the per-model scorecard above for the specific strengths and weaknesses on each dimension.

Task: High-quality anime-style illustration.

Image Generation5 modelsLast run: 2026-09-25

🏆 Winner

Midjourney v6.1 (--niji mode)

🥈 Runner-up

NovelAI v3

💰 Price-performance

Stable Diffusion 3.5 + anime LoRA (free local run)

For anime specifically, Midjourney's --niji mode is the clear winner — it was trained with anime-specific weighting. NovelAI v3 is the dedicated specialist and only marginally behind. Avoid FLUX and DALL-E for anime; they produce "photo-painted" output that lacks the hand-drawn feel.

The prompt (identical across all runs)

Copy-paste prompt

Anime illustration, young woman with silver hair wearing a kimono, standing on a rooftop at sunset overlooking Tokyo, dramatic composition, Studio Ghibli meets Makoto Shinkai aesthetic, hand-painted feel, soft lighting, highly detailed, 4:5.

Leaderboard

Rank	Model	Total /50	Cost / run
🥇	Midjourney v6.1 niji	49	4.0¢
🥈	NovelAI v3	47	2.0¢
🥉	Stable Diffusion 3.5 + anime LoRA	43	1.0‰ cent
#4	DALL-E 3	37	4.0¢
#5	FLUX 1.1 Pro	36	5.0¢

FLUX is a photorealism model — anime output felt like "anime filter on a photo" rather than hand-drawn. Pass.

Methodology

• Each model ran the exact same prompt 3-5 times (temperature 0.7 where configurable).
• We scored each run across 5 dimensions on a 1-10 scale, then averaged for the final.
• Cost is per completed run, rounded to the nearest 0.1¢.
• Last run: 2026-09-25. We re-run benchmarks when a model receives a major version bump.
• These are editorial benchmarks, not an academic study. We publish the exact prompt so you can reproduce.

More image generation benchmarks

Image — Cinematic portrait

Photorealistic cinematic portrait with correct hands and faces.

Image — Logo mark

A clean, scalable logo mark that reads at any size.

Image — Product photo

Studio-grade product photo for Shopify / Amazon.

FAQs — Image — Anime

Anime illustration, young woman with silver hair wearing a kimono, standing on a rooftop at sunset overlooking Tokyo, dramatic composition, Studio Ghibli meets Makoto Shinkai aesthetic, hand-painted feel, soft lighting, highly detailed, 4:5.

Rank

Model

Total /50

Cost / run

🥇

Midjourney v6.1 niji

4.0¢

🥈

NovelAI v3

2.0¢

🥉

Stable Diffusion 3.5 + anime LoRA

1.0‰ cent

DALL-E 3

4.0¢

FLUX 1.1 Pro

5.0¢

Per-model scorecard

🥇 Midjourney v6.1 niji

49 / 50 · 4.0¢ per run

Anime fidelity

10/10

Composition

10/10

Detail richness

10/10

Prompt adherence

9/10

Style consistency

10/10

niji mode is explicitly tuned for anime. Captured the Shinkai light-leak aesthetic perfectly.

NovelAI v3

47 / 50 · 2.0¢ per run

Anime fidelity

10/10

Composition

9/10

Detail richness

9/10

Prompt adherence

9/10

Style consistency

10/10

Dedicated anime model. Best style consistency across regens. UI is clunky.

Stable Diffusion 3.5 + anime LoRA

43 / 50 · 1.0‰ cent per run

Anime fidelity

9/10

Composition

8/10

Detail richness

9/10

Prompt adherence

8/10

Style consistency

9/10

Needed the Counterfeit or Anything v5 LoRA to get there. Free if self-hosted.

DALL-E 3

37 / 50 · 4.0¢ per run

Anime fidelity

6/10

Composition

9/10

Detail richness

7/10

Prompt adherence

9/10

Style consistency

6/10

Decent prompt adherence but the anime style was generic — reads more as "digital art" than "anime".

FLUX 1.1 Pro

36 / 50 · 5.0¢ per run

Anime fidelity

6/10

Composition

8/10

Detail richness

8/10

Prompt adherence

8/10

Style consistency

6/10

FLUX is a photorealism model — anime output felt like "anime filter on a photo" rather than hand-drawn. Pass.

Methodology

• Each model ran the exact same prompt 3-5 times (temperature 0.7 where configurable).

• We scored each run across 5 dimensions on a 1-10 scale, then averaged for the final.

• Cost is per completed run, rounded to the nearest 0.1¢.

• Last run: 2026-09-25. We re-run benchmarks when a model receives a major version bump.

• These are editorial benchmarks, not an academic study. We publish the exact prompt so you can reproduce.