SandboxThe sandbox directory
Every model and agent that runs inside Sandbox — what each one is best at, and a full guide on how to use it in the workspace. Click any card to read the guide.
ChatGPT · GPT-5
The generalist benchmark: writing, reasoning, plugins, voice. In Sandbox: parallel mode, o3 deep reasoning, custom GPTs imported.
Claude Opus & Sonnet
Long-document analysis, nuanced writing, agentic coding. In Sandbox: 200K context windows and Artifacts rendered live.
Gemini 2.5 Pro
Multimodal at massive scale — video, audio and million-token context. In Sandbox: Deep Research runs as a delegated agent.
Grok 4
Real-time awareness of what's happening now, unfiltered tone. In Sandbox: live social and news monitoring agents.
Midjourney v7
The art-direction gold standard — style references, consistent characters, print-grade output. Full commercial license included.
DALL·E 3 & Imagen
Prompt-faithful image generation with text rendering that actually works. Ideal for ads, UI mockups and diagrams.
Stable Diffusion 3
Open weights on our GPU fleet — LoRA fine-tunes, ControlNet pose control, and your custom styles stay private.
ElevenLabs
Most realistic voices and voice cloning on the market, 32 languages. In Sandbox: script → narration → final audio in one step.
Runway Gen-4
Text-to-video and video-to-video at broadcast quality. In Sandbox: storyboard with GPT-5, render with Runway, one credit pool.
Suno v4
Full songs with vocals from a text description — jingles, beds, theme music. Commercial rights on paid plans.
DeepSeek V3
Frontier-level reasoning at a fraction of the credit cost — the workhorse for high-volume tasks and batch processing.
Llama 4
Open-weights flagship, hosted on our clusters — your prompts never leave our infrastructure. Great for private workloads.
Mistral Large
European flagship with EU-friendly data terms — strong multilingual reasoning, French/German/ES quality that punches up.
Perplexity
Live web research with citations on every claim. In Sandbox: research briefs run as agents and export as sourced reports.
Code Agent
Autonomous coding: point it at a repo, describe the change, review the diff. Routes each step to Claude, GPT-5 or DeepSeek.
Research Agent
Market teardowns, competitor scans, due diligence packs — delegated once, delivered with sources and a cost receipt.
Media Agent
One brief in → script, voiceover, visuals and a rendered video out. Chains GPT-5, Midjourney, ElevenLabs and Runway itself.
Copy Agent
Launch emails, landing pages, ad sets — drafted across 4 models, ranked for your brand voice, localized to 29 languages.
Every card above. One subscription.
Turn on the models you need, compare them side by side, and pay one bill. No card, no email — 3 days free.