open comparison layer
Compare AI model pricing, benchmarks, and capabilities
Compare models on what actually matters — cost, billing model, architecture, and capability — from one honest, sourced registry. Built for developers and AI-native teams choosing between models.
34 models across 17 providers · Browse all image model pricing →
7 image benchmarks tracked · Updated September 9, 2026 · Reviewed September 9, 2026
69 models across 25 providers · Browse all language model pricing →
5 language benchmarks tracked · Updated September 7, 2026 · Reviewed September 9, 2026
30 models across 11 providers · Browse all video model pricing →
3 video benchmarks tracked · Updated September 9, 2026 · Reviewed September 9, 2026
31 models across 18 providers · Browse all audio model pricing →
Updated September 7, 2026 · Reviewed September 9, 2026
8 models across 10 providers · Browse all 3D model pricing →
Updated September 3, 2026 · Reviewed September 9, 2026
"My data story goes back further than this project — superannuation reporting, then eleven years at a Melbourne tech company, always with the same focus: serving data that had utility, not just data that existed. Modelglass is that same discipline applied to AI model pricing."
— Scott, founder · Read the full story on Medium →
Are you paying too much for AI?
It's hard to know. AI model information is fragmented. Pricing changes. Providers use different billing models. Benchmark scores are quoted without context. Modelglass puts that information into one structured, sourced registry.
Compare
Prices, capabilities and architectures side by side.
Verify
Know where the data came from and when it was checked.
Integrate
Use the registry through the API and MCP.
Track
See how models and pricing change over time.
Built for
For developers
Compare API pricing, context windows, capabilities and provider availability.
For AI engineers
Use structured model metadata through the API and MCP server.
For teams
Compare models before committing to an inference provider.
For researchers
Inspect benchmark scores and their provenance.
For agents
Query current model pricing and capability data programmatically.
Compare AI models by modality
Image generation models
View all image generation models →Adobe Firefly Image 3
Adobe (Firefly Services)
$0.020 / image
Animagine XL 3.1
Replicate
$0.0037 / image
DALL·E 3
OpenAI
from $0.040 / image
FLUX 1.1 [pro]
fal.ai, Replicate
from $0.040 / MP
FLUX.1 [dev]
Replicate, Together AI
from $0.025 / MP
FLUX.1 [pro]
Replicate
$0.055 / image
FLUX.1 [schnell]
fal.ai, Replicate
from $0.0030 / MP
FLUX.1 Kontext
Black Forest Labs
from $0.040 / image
Gemini 2.5 Flash Image
Google Cloud (Vertex AI)
$0.039 / image
Gemini 3 Pro Image
Google Cloud (Vertex AI)
from $0.13 / image
Gemini 3.1 Flash Image
Google Cloud (Vertex AI)
from $0.045 / image
GPT Image 1
OpenAI
from $0.011 / image
GPT Image 2
OpenAI
from $0.0060 / image
Hunyuan-DiT v1.1 (Distilled)
Replicate
$0.00097 / second
HunyuanImage 3.0
fal.ai
$0.10 / MP
Ideogram 2.0
Ideogram
from $0.050 / image
Ideogram 3.0
Ideogram
from $0.060 / image
Imagen 4
Google Cloud (Vertex AI)
no pricing
Krea-2-Turbo
fal.ai
$0.0080 / MP
Leonardo Phoenix
Leonardo.Ai
$0.0026 / credit credit‑based
Luma Uni-1.1
Luma AI
from $0.040 / image
Midjourney v7
Midjourney
from $10.00 / mo subscription
Recraft V3
Recraft
from $0.040 / image
Runway Gen-4 Image
Runway
$0.010 / credit credit‑based
Seedream 4.0
BytePlus (ModelArk)
$0.030 / image
Seedream 4.5
BytePlus (ModelArk)
$0.040 / image
Seedream 5.0 Pro
BytePlus (ModelArk)
from $0.045 / image
Stable Diffusion 3.5 Large
Replicate
$0.065 / image
Stable Diffusion 3.5 Large Turbo
Replicate
$0.040 / image
Stable Diffusion 3.5 Medium
Replicate
$0.035 / image
Stable Diffusion XL 1.0
Replicate
$0.0014 / second
Stable Image Core
Stability AI
$0.030 / image
Stable Image Ultra
Stability AI
$0.080 / image
Z-Image-Turbo
SiliconFlow
$0.0050 / image
Language models
View all language models →Anthropic
from $0.80 / 1M in
Anthropic
from $3.00 / 1M in
Anthropic
from $10.00 / 1M in
Anthropic
from $1.00 / 1M in
Anthropic
from $5.00 / 1M in
Anthropic
from $5.00 / 1M in
Anthropic
from $5.00 / 1M in
Anthropic
from $3.00 / 1M in
Anthropic
from $3.00 / 1M in
Anthropic
from $2.00 / 1M in
Cohere
from $2.50 / 1M in
Cohere
from $3.00 / 1M in
DeepSeek
no pricing
DeepSeek
no pricing
DeepSeek
from $0.28 / 1M in
DeepSeek
from $0.14 / 1M in
DeepSeek
from $0.43 / 1M in
Baidu
from $0.59 / 1M in
Google DeepMind
from $0.10 / 1M in
Google DeepMind
from $0.30 / 1M in
Google DeepMind
from $1.25 / 1M in
Google DeepMind
from $2.00 / 1M in
Google DeepMind
from $1.50 / 1M in
Zhipu AI
from $1.40 / 1M in
Zhipu AI
from $1.40 / 1M in
OpenAI
from $2.50 / 1M in
OpenAI
from $0.15 / 1M in
OpenAI
from $1.75 / 1M in
OpenAI
from $1.75 / 1M in
OpenAI
from $1.75 / 1M in
OpenAI
from $0.75 / 1M in
OpenAI
from $5.00 / 1M in
OpenAI
from $15.00 / 1M in
OpenAI
from $1.00 / 1M in
OpenAI
from $5.00 / 1M in
OpenAI
from $2.50 / 1M in
OpenRouter
from $0.017 / 1M in
xAI
no pricing
OpenRouter
from $0.13 / 1M in
Thinking Machines Lab
from $1.87 / 1M in
AI21 Labs
from $2.00 / 1M in
AI21 Labs
from $0.20 / 1M in
FriendliAI
no pricing
Moonshot AI
from $0.60 / 1M in
Moonshot AI
from $0.95 / 1M in
Moonshot AI
from $0.95 / 1M in
Moonshot AI
from $3.00 / 1M in
Mistral AI
$0.000 / 1M tok
OpenRouter
from $0.021 / 1M in
Groq, Together AI
from $0.59 / 1M in
Together AI
from $0.27 / 1M in
Together AI
from $0.18 / 1M in
Xiaomi
from $0.41 / 1M in
MiniMax
from $0.30 / 1M in
Mistral AI
from $0.50 / 1M in
Mistral AI
from $0.10 / 1M in
Amazon
from $0.30 / 1M in
OpenAI
from $2.00 / 1M in
OpenAI
from $1.10 / 1M in
OpenRouter
from $0.15 / 1M in
Together AI
from $1.20 / 1M in
OpenRouter
from $0.46 / 1M in
Reka AI
from $0.80 / 1M in
Upstage
from $0.15 / 1M in
Perplexity
from $1.00 / 1M in
Perplexity
from $2.00 / 1M in
Perplexity
from $3.00 / 1M in
Perplexity
from $2.00 / 1M in
StepFun
from $0.10 / 1M in
Video generation models
View all video generation models →Runway
$0.050 / second
Runway
$0.010 / credit
fal.ai
$0.20 / clip
Black Forest Labs
from $0.060 / second
Google DeepMind (Gemini API)
$0.10 / second
Runway
no pricing
Runway
$0.010 / credit
Runway
$0.010 / credit
$0.080 / second
fal.ai, MiniMax
from $0.076 / clip
Runway
from $0.15 / second
fal.ai
$0.40 / clip
Kling AI (Kuaishou)
$0.040 / second
Kling AI (Kuaishou)
$0.060 / second
Replicate
$0.048 / clip
fal.ai
from $0.040 / second
Replicate
$0.42 / clip
fal.ai
from $0.20 / clip
Luma AI
from $0.30 / clip
Runway
from $0.16 / second
BytePlus (ModelArk)
no pricing
BytePlus (ModelArk)
no pricing
BytePlus (ModelArk)
no pricing
OpenAI
from $0.10 / second
Google DeepMind (Gemini API)
$0.35 / second
Google DeepMind (Gemini API)
from $0.10 / second
Google DeepMind (Gemini API)
from $0.050 / second
fal.ai
from $0.035 / second
Replicate
from $0.090 / second
fal.ai
$0.050 / second
Audio models
View all audio models →3D generation models
View all 3D generation models →fal, Hi3D (Hitem3D)
from $0.020 / credit
Atlas Cloud, fal, Tencent Cloud, WaveSpeedAI, Replicate
from $0.022 / gen
Meshy
from $0.020 / credit
fal, Hyper3D
from $0.40 / gen
Stability AI
$0.010 / credit
Stable Point Aware 3D (SPAR3D)
Stability AI
$0.010 / credit
fal
from $0.020 / gen
Tripo AI
from $0.010 / credit
Recently updated
Pricing most recently re-verified against the provider's source.
- Hunyuan-DiT v1.1 (Distilled) Replicateverified 2026-08-22
- Animagine XL 3.1 Replicateverified 2026-08-14
- FLUX.1 Kontext Black Forest Labsverified 2026-07-23
Recently updated
Pricing most recently re-verified against the provider's source.
- o3verified 2026-08-29
- MiMo-V2.5-Proverified 2026-08-25
- Claude Sonnet 5verified 2026-08-14
Recently updated
Pricing most recently re-verified against the provider's source.
- FLUX 3 Videoverified 2026-08-06
- LTX-2.3verified 2026-07-23
- Seedance 2.0verified 2026-07-22
Recently updated
Pricing most recently re-verified against the provider's source.
- Azure Neural TTSverified 2026-08-31
- MAI-Transcribeverified 2026-08-31
- Grok Voice Think Fast 2.0verified 2026-08-10
Recently updated
Pricing most recently re-verified against the provider's source.
- Hunyuan3Dverified 2026-09-04
- Meshy 7verified 2026-09-04
- Rodinverified 2026-09-03
Capability benchmarks
sourced, citable scoresSWE-bench Pro, Aider Polyglot, Terminal-Bench — real-world coding task completion and agentic performance.
HLE (Humanity's Last Exam) — graduate-level reasoning across 100+ academic disciplines, no-tools and with-tools.
BrowseComp, OSWorld-Verified — persistent web research and live computer-use task completion.
GDPval-AA v2 — real-world knowledge work (documents, spreadsheets, slides) scored against a human-expert baseline.
The data layer for model routing.
Modelglass doesn't route your traffic — it keeps your router honest. It's a provenance-tracked feed of current model pricing and independently reported benchmark scores, served over REST and MCP, so the router you build can pick the cheapest model that actually clears the quality bar. Prices are append-only with a source on every entry: when a provider changes a price, your routing table finds out — not three weeks later.
Proof, not promise: the Modelglass Cost-Aware Router — a VS Code extension live on the Marketplace — already does exactly this, picking the cheapest model that clears the right benchmark bar for your task.