info DEMO DATA: scores shown are template sample values pending independent evaluation · as of 2026-09-05 · Methodology

INDEX / IMAGE_SYNTHESIS / MIDJOURNEY v6.1-PROD US-WEST-EVAL-02

SHA-256: e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855

Midjourney (v6.1)

Production Standard

Apex Text-to-Image Diffusion & Aesthetic Coherence Engine

V6.1 ARCHITECTURE WEB WORKSPACE V2 CHARACTER WEIGHT (--CW) RETEXTURE PIPELINE NIJI V6 CORE

Airecmark Composite

96.2 / 100

Visual Synthesis

Launch Midjourney Web arrow_outward

Aesthetic Coherence auto_awesome

98.7%

Industry apex for photorealism, subsurface epidermal scattering, and volumetric lighting accuracy.

Text In-Image Precision spellcheck

92.4%

High precision rendering of typographic quotations and glyph layout inside raster boundaries.

Character Consistency face_6

94.6%

--cref & --cw facial bone structure and wardrobe locking.

P95 Fast GPU Latency timer

18.5s

Full 4-grid generation across unified H100/A100 inference infrastructure with immediate parallel previews.

Empirical Intelligence Vectors

Evaluated on 50,000 blind cross-prompt runs against human & automated judges

N=50k Eval

Micro-Texture Fidelity (Epidermal, weave, specular caustics) 99.1%

Complex Compositional Lighting (Raytrace coherence & bounce) 97.8%

Stylistic Versatility (Analog 35mm film to anime/hyper-stylized) 98.0%

Inpainting & Region Variant Consistency 95.0%

Prompt Syntactic Adherence (Semantic binding) 93.4%

Standard Deviation: ±1.42% across test clusters

mj-eval-telemetry.sh

LIVE INFERENCE

# Deterministic Inference Specifier

$ midjourney-cli run --prompt \

"Cinematic portrait of an orbital cybernetics engineer, natural volumetric dusk lighting, 70mm lens, photorealistic macro skin texture" \

--v 6.1 [v6.1 8-bit Latent Model]

--ar 16:9 [Native Aspect Ratio]

--style raw [Neutral Hyper-Real Bias]

--s 250 [Stylize Amplitude (0-1000)]

--cw 100 [Full Attire & Facial Weight]

--cref <id> [Cross-Attention Embedding]

Compute: Unified Matrix NV-H100 STATUS: READY (0.12s)

Model Output Artifact Inspection

Evaluation of high-frequency textures, depth plane separation, and specular lighting in native v6.1 generation.

Zoom: 100% Native Color Profile: DCI-P3

Prompt: Photoreal Specular Epidermis

Fidelity: 99.4%

Prompt: Caustic Volumetric Glass

Reflections: 98.6%

Prompt: Typography & Composition

Text Glyph: 94.1%

Quantitative Technical Specification

Core Engine Components

filter_vintage

8-Bit Latent Precision & Texture Upscaler

Completely refactored latent decoder pipeline eliminating chromatic banding in low-contrast zones. Native micro-detail synthesis handles hair, skin pores, and woven cloth with deterministic micro-reflections.

Latent Depth: 8-Bit Perceptual Banding Reduction: 98%

aspect_ratio

Canvas Pan, Zoom & Inpainting Suite

The Web Workspace V2 provides continuous canvas manipulation with fractional pan/zoom (1.5x, 2.0x, custom aspect ratios) paired with semantic region inpainting directly inside the web browser client.

Web Workspace: Native V2 Infinite Edge Synthesis

fingerprint

Style Reference & Character Lock

Multi-modal cross-attention conditioning allows users to feed image URLs into --sref for color/style transfer and --cref with customizable character weight (--cw 0-100).

Vectors: --sref / --cref / --sref random Facial Lock: 94.6%

texture

Retexture & Geometry Invariance

The latest Web Editor feature enables geometry invariance: re-render entirely novel textures, materials, and lighting setups over existing camera angles and structural silhouettes without shifting perspective coordinates.

Depth Preservation: 99.2% Perspective Drift: <0.05%

Direct Peer Comparison Matrix

Empirical score comparisons normalized across top generative diffusion platforms.

Updated: Today 06:00 UTC

Model / Architecture	Composite Score	Aesthetic Rank	Prompt Adherence	Typographic Purity	Pricing Model	Deployment Surface	Action
Midjourney v6.1 State of the Art (Aesthetic Apex)	96.2	#1 (98.7%)	93.4%	92.4%	$10 - $120/mo	Web Workspace + Discord	CURRENT
Flux.1 [dev/schnell] Black Forest Labs (Flow Matching)	95.0	#2 (96.4%)	97.8%	94.8%	Open Weights / API	Self-hosted / Cloud API
Ideogram 2.0 Ideogram Inc. (Graphic Typography)	93.5	#4 (91.2%)	95.1%	98.5%	$8 - $48/mo	Web Interface + API
Stable Diffusion 3.5 Large Stability AI (MMDiT 8.1B)	91.2	#5 (90.4%)	94.2%	91.0%	Free Community / Lic	Local ComfyUI / API
DALL-E 3 OpenAI (Diffusion with GPT-4 Rewriter)	90.8	#6 (88.1%)	96.0%	93.0%	$20/mo (ChatGPT Plus)	ChatGPT Web / OpenAI API

Commercial Tiers & Compute Allocation

Deterministic fast GPU hours, concurrent queue allocation, and license terms.

Entry Tier

Basic Plan

$10 / month

check 3.3 hrs/mo Fast GPU time
check ~200 image generation runs
check Commercial usage license
close No Relaxed unlimited GPU

Standard Plan

$30 / month

check 15 hrs/mo Fast GPU time
check Unlimited Relaxed GPU hours
check Web Workspace Canvas access
check 3 concurrent Fast queue jobs

Professional

Pro Plan

$60 / month

check 30 hrs/mo Fast GPU time
check Stealth Mode (Private generations)
check 12 concurrent fast jobs
check Unlimited Relaxed generations

Agency & Studio

Mega Plan

$120 / month

check 60 hrs/mo Fast GPU time
check Stealth Mode included
check Maximal concurrent throughput
check Priority queue routing

workspace_premium

Analyst Consensus Verdict

Unanimous Pass

Midjourney v6.1 continues to dominate the visual synthesis paradigm. While open alternatives like Flux.1 excel in prompt instruction adherence and Ideogram leads complex text placement, Midjourney preserves an unassailable aesthetic authority in lighting volumetrics, micro-skin specular realism, and editorial composition.

Merkle Root: 77f98b...9c32aa Signature: ED25519-POLY1305 Audit Sequence ID: #MJ-61-90822-PROD