inference.sh
One API key and CLI for 150+ AI models and apps: image, video, LLMs, search, 3D, and audio.
About
inference.sh is a unified platform for running AI models without managing infrastructure. One account and API key covers image generation (FLUX, Seedream, Grok Imagine), video (Veo, Wan, Seedance), LLMs (Claude, Gemini, Kimi, plus any OpenRouter model), search (Tavily, Exa), 3D generation, and audio, alongside utilities like media merging and X/Twitter automation. An infsh CLI wraps it all: install with a shell script, log in once, then run any app with infsh app run and JSON input.
The vendor's FAQ describes skills that improve with use, an open-source runtime, BYOK GPU support for self-hosting, and a shared context layer for teams; treat the open-source and self-host claims as vendor-stated since the public repo footprint is thin. It is a young aggregator in a crowded field of model routers, so the honest question is whether its catalog depth and reliability beat going direct to the underlying providers or to an established router.
What it can do
- Single API key across 150+ AI models and apps
- infsh CLI for running any model from the terminal
- Image, video, LLM, search, 3D, and audio generation
- Vendor-stated open-source runtime with BYOK self-hosting
- Team workspaces with shared context layer
Setup
- Install the CLI:
curl -fsSL https://cli.inference.sh | sh. - Create an account at inference.sh and run
infsh login. - List available apps:
infsh app list --search "flux". - Run one:
infsh app run <app> --input '{"prompt": "..."}'.
Prerequisites
An inference.sh account; payment method for usage beyond free limits.
Use cases
Works with
Trust score — 5/10
Scored 5/10 (metrics tier, metrics-v1). Evidence: inference.sh verified live with current product FAQ; referenced by multiple third-party agent-skill repos on GitHub (kevinnft/ai-agent-skills, learn-skills.dev); infsh CLI install path documented. Strengths: broad model catalog behind one key, clear CLI UX. Docked for: (1) no public adoption metrics, funding, or team information to verify; (2) open-source runtime and self-host claims are vendor-stated only. License: proprietary (vendor claims open-source runtime).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 4.0/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 4.0/10
- Manual review · sanity read + red-flag check
- 5.0/10
- · inference.sh verified live with current product FAQ
- · Referenced by third-party agent-skill repos: kevinnft/ai-agent-skills, neversight/learn-skills.dev
- · infsh CLI install path documented (cli.inference.sh)
- · No public funding, team, or adoption metrics found
- · Open-source runtime and BYOK self-host are vendor-stated only
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.