# legit2compute > legit2compute is a community compute network: people lend their NVIDIA graphics cards' idle hours to AI and media jobs and are credited for every checked job. Developers use the same network through an OpenAI-compatible API for open-weight language and vision models, with music, voice, image and video generation on the same key. Pay as you go in credits (1 credit = 1 US cent), no subscription. - API: keep the OpenAI SDK you use and change the base URL. Per-token prices for every model are on the models page. - `l2c/auto` as the model routes each request to the cheapest model that fits it. - Images and video: text to image, image edits, text to video and image to video are jobs on the same key (submit, poll, download), priced per result: from $0.004 an image and $0.07 per second of video (a 5-second clip with sound is $0.35). - Prompts and outputs are not stored. A Private switch on the developer dashboard keeps an account's model requests on machines the operator owns, never a community provider's or an outside service. - Agents: setup guides for Hermes Agent, OpenClaw and Odysseus, models with tool calling and up to 256K context, and a public demo key in the guides to try one always-on model before signing up. - MCP: a hosted Model Context Protocol server at https://legit2compute.com/mcp (streamable HTTP, your API key as a Bearer header), listed in the official MCP Registry as com.legit2compute/mcp. Claude Code, Claude Desktop, Cursor and VS Code can ask the models and make images, music, voice and video through it. - Model hosting by providers is not open yet; owners of NVIDIA cards with 16 GB or more can join a waitlist. - Card owners: every paid job a provider's card finishes is checked and credited. Letting the card host free users' demos for part of its online time (unpaid, the provider's choice) raises the provider's share of every paid job and gives it more turns at paid work. See /earn-with-your-gpu, /gpu-mining-alternative and /rent-out-your-gpu. - Providers: a small agent for Windows or Linux PCs with an NVIDIA graphics card lends the GPU's idle hours, steps aside when you're back, and credits each checked job. Almost all paid work needs the GPU. - A free chat demo needs no account. ## Docs - [Full LLM reference (live models & pricing)](https://www.legit2compute.com/llms-full.txt) - [RSS updates feed](https://www.legit2compute.com/feed.xml) - [Earn with your GPU](https://www.legit2compute.com/earn-with-your-gpu): what an NVIDIA graphics card needs, interactive VRAM checker, how providers are paid, away mode (also in [Español](https://www.legit2compute.com/ganar-con-tu-gpu), [Português](https://www.legit2compute.com/ganhar-com-sua-gpu), [Deutsch](https://www.legit2compute.com/mit-gpu-geld-verdienen), [Français](https://www.legit2compute.com/gagner-avec-son-gpu), [日本語](https://www.legit2compute.com/gpu-de-kasegu), [한국어](https://www.legit2compute.com/gpu-suik-changchul), [Türkçe](https://www.legit2compute.com/gpu-ile-para-kazan), [Indonesia](https://www.legit2compute.com/hasilkan-uang-dengan-gpu)) - GPU hardware guides: [RTX 3060 12GB](https://www.legit2compute.com/earn-with-rtx-3060), [RTX 3090 24GB](https://www.legit2compute.com/earn-with-rtx-3090), [RTX 4070 / Ti](https://www.legit2compute.com/earn-with-rtx-4070), [RTX 4080 16GB](https://www.legit2compute.com/earn-with-rtx-4080), [RTX 4090 24GB](https://www.legit2compute.com/earn-with-rtx-4090), [RTX 5090 32GB](https://www.legit2compute.com/earn-with-rtx-5090), [48 GB RTX A6000 / 6000 Ada](https://www.legit2compute.com/earn-with-rtx-a6000), [RTX PRO 6000 96GB](https://www.legit2compute.com/earn-with-rtx-pro-6000), [DGX Spark](https://www.legit2compute.com/earn-with-dgx-spark) - [Salad alternative](https://www.legit2compute.com/salad-alternative): idle NVIDIA GPU sharing with real AI jobs, away-mode yielding, free self-clipping on your own card, and PayPal/bank/crypto or credit bonus payouts - [What happened to ETH & crypto GPU mining?](https://www.legit2compute.com/what-happened-to-eth-mining): why BTC moved to ASICs, ETH mining ended in 2022, and distributed AI compute for everyday tasks replaced hash mining for home NVIDIA GPUs - [Distributed AI compute & FOSS models](https://www.legit2compute.com/distributed-ai-compute): how one shared compute layer powers legit2clip, Synestia, legit2cloud, legit2click and an OpenAI-compatible API running live open-weight FOSS models - [Vast.ai alternative](https://www.legit2compute.com/vast-ai-alternative): single-job containerized provider nodes without full-machine rentals, plus pay-per-token / pay-per-result APIs without idle VM or storage bills - [OpenAI-compatible API](https://www.legit2compute.com/openai-compatible-api): base URL, key, live model prices and l2c/auto in one page - [Private LLM API for AI agents](https://www.legit2compute.com/private-llm-api-for-agents): the Private switch, which models suit agents, and a demo key to try - [Host models on your GPU](https://www.legit2compute.com/host-models): the waitlist for NVIDIA cards with 16 GB or more - [Developers](https://www.legit2compute.com/developers): API keys and getting started - [Models and prices](https://www.legit2compute.com/models): every model, its capabilities and per-token price - [Live benchmarks](https://www.legit2compute.com/benchmarks): median response time, output tokens per second and answer rate per model from real traffic, with a free JSON feed at https://www.legit2compute.com/benchmarks.json - [Image and video API](https://www.legit2compute.com/media): what it makes, prices per result, real samples and the three calls - [Agent skill](https://www.legit2compute.com/agent-skill): one download that teaches Claude Code, Hermes Agent or OpenClaw to make video Shorts, music, voiceovers, sound effects and images with your key; real samples and setup steps - [Sora API alternative](https://www.legit2compute.com/sora-api-alternative): text to video and image to video with sound by API, for developers who used the Sora API (discontinued 2026-09-24) - [Text-to-video API](https://www.legit2compute.com/text-to-video-api): generate 5–15s MP4 clips with synced sound for $0.07/sec, pay as you go - [Image-to-video API](https://www.legit2compute.com/image-to-video-api): animate photos and artwork with synced audio for $0.07/sec - [Runway API alternative](https://www.legit2compute.com/runway-api-alternative): pay-per-second video API with sound included and no monthly minimum - [Kling API alternative](https://www.legit2compute.com/kling-api-alternative): direct pay-as-you-go video API with synced audio and optional Private mode - [Video & media API cost calculator](https://www.legit2compute.com/video-api-cost-calculator): interactive monthly spend estimator for video, images, voice and music - [Cheapest AI video APIs compared](https://www.legit2compute.com/cheapest-ai-video-apis-compared): five video generation APIs (legit2compute, Runway, Kling, Luma, fal.ai/Replicate) compared side by side on per-second cost, built-in sound and refunds - [One-key AI media API](https://www.legit2compute.com/one-key-ai-media-api): one API key for open-weight LLMs, video with sound, images, voiceovers, SFX and music - [Agent quickstart](https://www.legit2compute.com/agent-quickstart.md): plain Markdown for a coding agent setting up a project: the key, the OpenAI-compatible base URL, media jobs and the errors to expect - [Docs](https://www.legit2compute.com/docs): API reference and integration guides - [Hermes Agent setup](https://www.legit2compute.com/docs/hermes-agent) - [OpenClaw setup](https://www.legit2compute.com/docs/openclaw) - [Odysseus setup](https://www.legit2compute.com/docs/odysseus) - [MCP server setup](https://www.legit2compute.com/docs/mcp): one URL and your key for Claude Code, Claude Desktop, Cursor and VS Code - [Network status](https://www.legit2compute.com/status) - [Chat demo](https://www.legit2compute.com/chat) ## Optional - [legit2clip](https://www.legit2clip.com/): long videos into captioned short clips, from the same team - [legit2cloud](https://www.legit2cloud.com/): AI music, voice, images and video on one account - [Synestia](https://synestia.bot/): AI music and a free song library for videos - [legit2click](https://legit2click.com/): AI-run news sites on your own domain, on the same account ## Live Public LLM Catalog (OpenAI-Compatible) Base URL: `https://api.legit2compute.com/nodes/llm/v1` Header: `Authorization: Bearer ` Auto-router model ID: `l2c/auto` - `gemma4:12b` ($0.040/1M in, $0.12/1M out, 128K ctx) - `granite3.2-vision:2b` ($0.040/1M in, $0.060/1M out, 16K ctx) - `minicpm-v4.6:latest` ($0.040/1M in, $0.060/1M out, 32K ctx) - `minicpm-v:latest` ($0.040/1M in, $0.060/1M out, 32K ctx) - `llama3.1` ($0.030/1M in, $0.050/1M out, 128K ctx) - `ornith-1.5:9b` ($0.080/1M in, $0.14/1M out, 256K ctx) - `ornith-1.5:35b` ($0.12/1M in, $0.80/1M out, 256K ctx) - `qwen3.8:27b` ($0.15/1M in, $1.80/1M out, 128K ctx) - `bonsai-2:9b` ($0.064/1M in, $0.10/1M out, 256K ctx) - `deepseek-ocr:3b` ($0.040/1M in, $0.060/1M out, 128K ctx) - `glm-ocr:latest` ($0.040/1M in, $0.060/1M out, 16K ctx) - `mimo-v2.6:9b` ($0.030/1M in, $0.10/1M out, 256K ctx) - `muse-glimmer:30b` ($0.25/1M in, $1.00/1M out, 32K ctx) - `bge-m3:latest` ($0.008/1M in, 8K ctx) - `granite4.2:3b` ($0.040/1M in, $0.060/1M out, 128K ctx) - `granite4.2:8b` ($0.050/1M in, $0.16/1M out, 128K ctx) - `lfm3:1.7b` ($0.040/1M in, $0.060/1M out, 256K ctx) - `qwen3-embedding:0.6b` ($0.005/1M in, 32K ctx) - `qwen3-embedding:8b` ($0.008/1M in, 32K ctx) - `qwen3-reranker:0.6b` ($0.030/1M in, 8K ctx) - `whisper-large-v3-turbo` ($8.33/1M in) ## Media Jobs API (Same Key, Pay Per Result) - Endpoint: `POST https://api.legit2compute.com/jobs` with header `x-api-key: ` - Poll: `GET https://api.legit2compute.com/jobs/` - Download: `GET https://api.legit2compute.com/jobs//download` - Pipelines: - `video_t2v` / `video_i2v` (5, 10 or 15s MP4 with synced sound): $0.07/sec - `image_generate` / `image_edit`: from $0.004/image - `voice_generate` / `moss_audio` (voiceovers & SFX): from $0.005–$0.04/clip - `music_generate`: from $0.08/track - Failed jobs are automatically refunded. Optional `"private": true` keeps execution strictly on operator-owned servers. - Payments: Cards, Apple Pay, Google Pay, Link, and supported local Stripe payment methods. No monthly subscription.