Language & vision models

An OpenAI-compatible API, on a crowd-powered GPU cloud.

Point any OpenAI client at our endpoint and use the models below — text, images in, tool calling, long context. Prompts and outputs are never stored; a private tier keeps your traffic on our own machines. Where the same model is sold elsewhere, ours is priced below the cheapest provider listed on OpenRouter.

Get an API keyDocs & integration guidesAPI reference ↗

Catalog · ● 4 of 9 resident now · the rest start on your first request

$ per 1M tokens · updated every 30 s
Model & capabilitiesInput / outputCached inputAvailability
Granite 3.2 Vision 2B
granite3.2-vision:2b T1
text image in 16K ctx
 $0.040 / $0.060 $0.020 ○ on demand
MiniCPM-V 2.6 8B (vision)
minicpm-v:latest T1
text image in 32K ctx
 $0.040 / $0.060 $0.020 ● available now
Llama 3.1 8B
llama3.1 T1
text tools 128K ctx
 $0.016 / $0.032 $0.008 ● available now
Ornith 1.5 9B (vision, 256K)
ornith-1.5:9b T2
text image in tools 256K ctx
 $0.050 / $0.080 $0.020 ● available now
Ornith 1.5 35B (vision, 256K)
ornith-1.5:35b T3
text image in tools 256K ctx
 $0.18 / $0.30 $0.050 ○ on demand
Qwen3.8 27B (vision + text)
qwen3.8:27b T3
text image in tools 128K ctx
 $0.12 / $1.50 $0.030 ● available now
Llama 3.1 70B
llama3.1:70b T3
text tools 128K ctx
 $0.32 / $0.32 $0.10 ○ on demand
DeepSeek-OCR 3B
deepseek-ocr:3b
text image in tools 128K ctx
 $0.040 / $0.060 $0.020 ○ on demand
Granite 4.1 3B (text)
granite4.1:3b
text tools 128K ctx
 $0.040 / $0.060 $0.020 ○ on demand

Send images as OpenAI image_url content parts (data URL or https) to any model marked image in. A model that is not resident answers 503 with "warming": true; retry in a minute or two. Every response names who served it in X-L2C-Provider.

Auto mode — l2c/auto

Send l2c/auto as the model and the gateway picks the cheapest resident model in the tier your request calls for, billing at that model's price. Cap or floor it per request with route: {"max_tier": n, "min_tier": n}; the response carries X-L2C-Routed-Model and X-L2C-Auto-Tier, and your Activity page shows what auto saved.

Tier 1 · small — chat, rewrites, summaries, simple questions
granite3.2-vision:2b ○  minicpm-v:latest ●  llama3.1 ●
Tier 2 · mid — structured output, code, tools, moderate analysis
ornith-1.5:9b ●
Tier 3 · large — complex reasoning, long documents, architecture, maths
ornith-1.5:35b ○  qwen3.8:27b ●  llama3.1:70b ○

Get access in a minute

Create an API key in the developer portal, keep the OpenAI SDK you already use, and change the base URL. Credits are shared with everything else on the platform — 1 credit = $0.01.

curl https://api.legit2compute.com/nodes/llm/v1/chat/completions \
  -H "x-api-key: $L2C_KEY" -H "content-type: application/json" \
  -d '{"model":"l2c/auto","messages":[{"role":"user","content":"Hello"}]}'
  • Usage, tokens and cost per request in your Activity page
  • Private inference tier for sensitive workloads
  • A share of every token billed goes to the GPU host that served it