NEWGPU owners: host free demos on your card and earn a bigger share of every paid job. See how →

Architecture · Open-source AI · Shared compute layer

Distributed AI compute for everyday tasks — powered by open-source models

The next era of personal GPUs isn’t crypto mining — it’s distributed AI compute for daily work. legit2compute is one unified compute layer that turns idle NVIDIA graphics cards into real AI infrastructure powering video clipping, music creation, voiceovers, image/video generation, automated newsrooms, and an OpenAI-compatible FOSS model API.

The short answer. One distributed compute network, one account, and one pay-as-you-go credit balance (1 credit = $0.01) run all five ecosystem apps — from clipping 12-hour streams on legit2clip and composing songs on Synestia to calling open-weight FOSS LLMs and $0.07/sec video-with-sound APIs.

One compute layer, five everyday products

Every workload uses the exact same job scheduler, verification layer, and credit wallet — so real user demand from creators, streamers, churches, and developers keeps distributed GPUs busy doing useful work:

Everyday productDaily task solvedWorkloads dispatched to the same compute layer
legit2clip.comTurn long YouTube, Twitch, Kick streams, podcasts & sermons into captioned ShortsSpeech transcription, highlight selection, vertical facecam reframing, caption burn-in & Studio renders
synestia.botCreate original songs, cover art & voiceovers in browser or Discord /createOpen-weight AI music synthesis, cover artwork generation, voiceovers & sound effects
legit2cloud.comOne-stop AI studio for everyday creators (video with sound, images, audio & chat)Text-to-video & image-to-video with synced audio, image generation/editing & chat
legit2click.comAI-run local/niche newsrooms & conference sites on custom domainsAutomated open-weight LLM drafting, editorial workflows & artwork on shared credits
legit2compute APIOpenAI-compatible LLM & media API for apps, scripts and AI coding agentsOpen-weight chat/vision/tool-calling models (/nodes/llm/v1) + media jobs (/jobs)

Live FOSS & open-weight models on the compute layer

Every text and vision model below is callable right now via standard OpenAI SDKs at https://api.legit2compute.com/nodes/llm/v1 (or via l2c/auto automatic cost routing), with zero monthly subscription:

Open-weight / FOSS modelTypeContextIn / Out per 1MStatus
Llama 3.1 8B
llama3.1
text 128K $0.030 / $0.050 ● live
MiniCPM-V 2.6 8B (vision)
minicpm-v:latest
vision 32K $0.040 / $0.060 ● live
Ornith 1.5 9B (vision, 256K)
ornith-1.5:9b
vision 256K $0.080 / $0.14 ● live
Ornith 1.5 35B (vision, 256K)
ornith-1.5:35b
vision 256K $0.12 / $0.80 ● live
BGE-M3
bge-m3:latest
embedding 8K $0.008 / — ○ on demand
Qwen3 Embedding 0.6B
qwen3-embedding:0.6b
embedding 32K $0.005 / — ○ on demand
Qwen3 Embedding 8B
qwen3-embedding:8b
embedding 32K $0.008 / — ○ on demand
Qwen3 Reranker 0.6B
qwen3-reranker:0.6b
rerank 8K $0.030 / — ○ on demand
Whisper large-v3 turbo
whisper-large-v3-turbo
transcription — $8.33 / — ○ on demand
Granite 3.2 Vision 2B
granite3.2-vision:2b
vision 16K $0.040 / $0.060 ○ on demand
MiniCPM-V 4.6 (vision)
minicpm-v4.6:latest
vision 32K $0.040 / $0.060 ○ on demand
DeepSeek-OCR 3B
deepseek-ocr:3b
vision 128K $0.040 / $0.060 ○ on demand
GLM-OCR
glm-ocr:latest
vision 16K $0.040 / $0.060 ○ on demand
Granite 4.2 3B (text)
granite4.2:3b
text 128K $0.040 / $0.060 ○ on demand

Pulled live from the public model catalog (all 21 models · live speed benchmarks).

Built for both sides of the flywheel

For creators & developers

  • One key & wallet: Open-weight LLMs, video with synced sound ($0.07/s), images (from $0.004), voiceovers, SFX, music, and stream clipping.
  • Zero monthly lock-in: Pay strictly per token or finished output; failed jobs auto-refund.
  • Optional Private mode: Keep sensitive or commercial prompts strictly on operator-owned servers with zero prompt storage.

For NVIDIA GPU owners

  • Real workloads, not hash waste: Your idle card renders actual clips, transcripts, audio and AI jobs inside isolated Docker containers.
  • Gaming-safe away mode: Yields instantly when you touch mouse/keyboard or launch a game.
  • Free self-clipping + live payout ladder: Your own legit2clip jobs run free on your GPU while online, and paid jobs credit your account (from 5% to 25%).

Questions

How does one compute layer power legit2clip, Synestia, legit2cloud, legit2click and the API?

Every app submits standardized jobs (video clipping, transcription, music, voice, images, video with sound, or open-weight LLM inference) into the same orchestrator and unified credit wallet. Community NVIDIA GPUs and operator servers execute isolated containers, verify outputs, and credit both the provider and user seamlessly.

Which open-source (FOSS) AI models are available?

Our live catalog of public open-weight chat, reasoning, vision, embedding and rerank models is updated continuously on /models (and shown in the live table on this page), alongside open-source media pipelines for transcription, captions, audio, images and video.

What if my business workload cannot run on community GPUs?

Flip the Private switch on your developer dashboard (or pass "private": true on media jobs) so every request runs strictly on operator-owned servers. Prompts and outputs are never stored.

How do everyday people participate with their home PC?

Install legit2node on Windows or Linux with an NVIDIA GPU. Away mode runs containerized jobs only while you are not using your PC, steps aside immediately when you game or work, runs your own legit2clip jobs on your GPU for free, and credits paid jobs on our live payout ladder (from 5% to 25%).

Explore further: what happened to ETH & crypto mining · full live FOSS model list · live speed benchmarks · one-key AI media API · earn with your GPU.

Put your graphics card to work while you're away.

Free to join, paused whenever you're using it, and a bigger share of every paid job when it hosts free demos.