Providers
Modelglass tracks the companies building and serving today's AI models — their pricing, their model lineups, and how those models perform.
242 models with tracked pricing across five model types — image, language, video, audio and 3D — from 57 providers. Frontier language models also carry capability-benchmark scores (coding, science, agentic, GDPval).
Adobe (Firefly Services)
1 model tracked
AI21 Labs
2 models tracked
Amazon Web Services
4 models tracked
Anthropic
54 models tracked
AssemblyAI
1 model tracked
Atlas Cloud
1 model tracked
Baidu
1 model tracked
Black Forest Labs
2 models tracked
BytePlus (ModelArk)
10 models tracked
Cartesia
1 model tracked
Cloudflare Workers AI
1 model tracked
Cohere
3 models tracked
Deepgram
1 model tracked
DeepSeek
14 models tracked
ElevenLabs
5 models tracked
fal.ai
15 models tracked
Fish Audio
4 models tracked
FriendliAI
1 model tracked
43 models tracked
Groq
3 models tracked
Hi3D (Hitem3D)
1 model tracked
Hyper3D
1 model tracked
Ideogram
2 models tracked
Inworld AI
3 models tracked
Kling AI (Kuaishou)
5 models tracked
Leonardo.Ai
1 model tracked
Luma AI
2 models tracked
Meshy
1 model tracked
Meta
9 models tracked
Microsoft Azure
3 models tracked
Midjourney
1 model tracked
MiniMax
7 models tracked
Mistral AI
12 models tracked
Moonshot AI
12 models tracked
OpenAI
59 models tracked
OpenRouter
5 models tracked
Perplexity
4 models tracked
PlayHT
1 model tracked
Qwen (Alibaba)
6 models tracked
Recraft
1 model tracked
Reka AI
1 model tracked
Replicate
14 models tracked
Runway
11 models tracked
SiliconFlow
1 model tracked
Stability AI
7 models tracked
StepFun
5 models tracked
Suno
1 model tracked
Tencent Cloud
1 model tracked
Thinking Machines Lab
1 model tracked
Together AI
5 models tracked
Tripo AI
1 model tracked
Udio
1 model tracked
Upstage
6 models tracked
WaveSpeedAI
1 model tracked
xAI
11 models tracked
Xiaomi
7 models tracked
Zhipu AI
9 models tracked
The model count is distinct models across the five priced verticals (image, language, video, audio, 3D), deduplicated so a model offered by more than one provider counts once; some of those models also carry capability-benchmark scores, which are attached to the same entry rather than counted again. Provider identity is joined across the image, LLM, video and audio provider records, plus the science, coding, agentic and GDPval model_id prefixes, via an explicit, human-reviewed alias list. See any provider's page for the full explanation. A provider added under a new, differently-spelled slug in each vertical will appear here as separate cards until that alias entry is added; this list reflects the registry as of the last build, not a live merge check.
Modelglass is not affiliated with or endorsed by the companies listed here.