The engine room
The models behind the Skills
A Skill is a folder of instructions. The model is what reads them, runs the tools and carries the work to the end. These are the engines the catalogue is written against.
14 models from 10 labs · 9 agentic, 5 video
Anthropic · Claude 5.5
Opus 5.5
Anthropic's newest frontier model, built to carry long, multi-step work from one prompt to a finished result.
- Context
- 1M tokens
- Best at
- Agentic coding
OpenAI · GPT-6
GPT-6 Astra
OpenAI's frontier model for end-to-end agentic work, reasoning across a task rather than a single turn.
- Context
- ≈1M tokens
- Best at
- End-to-end work
Anthropic · Claude Fable 5.1
Fable 5.1
Anthropic's top model for demanding reasoning and long-horizon agentic work, with adaptive thinking.
- Context
- 1M tokens
- Best at
- Demanding reasoning
Anthropic · Claude Opus 5
Opus 5
Anthropic's Opus-tier flagship: near-frontier intelligence for complex coding and enterprise work.
- Context
- 1M tokens
- Best at
- Complex coding
Moonshot AI · Kimi K3
Kimi K3
Moonshot AI's natively multimodal frontier model, top-ranked for front-end web development.
- Context
- 1M tokens
- Best at
- Front-end craft
xAI · Grok 4
Grok 4.6
A 500K-token context and stronger long-running agents, built for ambitious interactive builds.
- Context
- 500K tokens
- Best at
- Interactive builds
Z.ai · GLM 5.3
GLM 5.3 Flash
One-shot sites, browser games and 3D scenes at flash speed and near-zero cost.
- Context
- 1M tokens
- Best at
- Fast, low-cost coding
DeepSeek · V4.1
DeepSeek V4.1 Flash
A natively multimodal 552B-parameter model built to read images as well as text, cheaply.
- Input
- Text and images
- Best at
- Cheap batch runs
Alibaba · Qwen 3.8
Qwen 3.8 Max
Alibaba's flagship Qwen model, a standout at one-shot 3D worlds, WebGL games and front-end builds.
- Best at
- 3D and games
- Strength
- Front-end builds
ByteDance · Seedance 2.5
Seedance 2.5
ByteDance's latest video model: a 30-second audio-video scene from one pass, with up to 50 references.
- Duration
- 30s, extendable
- References
- 50 assets
ByteDance · Seedance 2.0
Seedance 2.0
ByteDance's unified audio-video model: text, image, audio and video references for 15-second multi-shot sequences.
- Duration
- 15s, multi-shot
- References
- 15 assets
Kuaishou · Kling 4.0
Kling 4.0
Kuaishou's production-grade model: 30-second takes, 10-bit HDR, omni-reference and precise keyframes.
- Duration
- 30s
- Output
- 4K HDR
Alibaba · Wan 3.0
Wan 3.0
Alibaba's all-in-one video model: 30 seconds, native sound, and references that include documents and web pages.
- Duration
- 30s at 30fps
- References
- Up to 20 assets
MiniMax · H3
MiniMax H3
MiniMax's open-source omni-modal model: 15-second clips, 2K output and native stereo audio.
- Duration
- 4–15s at 24fps
- Licence
- Open weights
Capabilities and specifications follow each lab's published profile. What your own agent offers depends on its plan, region and configuration.
How to choose
Start with the work, not the leaderboard. A long build that touches many files wants context and stamina; a page that has to look designed wants front-end taste; a video wants a different kind of model entirely.
The same Skill can run on any of them. Nothing in the catalogue is tied to one model — the folders are plain instructions, and your agent brings the engine. Choosing per task, rather than per install, is the point.
Volume changes the arithmetic. When a Skill runs on every ticket, a fast sparse model is often the better default, with a frontier model held back for the run that has to be right the first time.
Ready to extend your agent?
Connect once, then install any Skill from our studio straight from your agent.