llms.txt
Compare with other sites' llms.txt files →# Scalattice
> Managed production inference across a distributed GPU network.
Developers call an OpenAI-compatible API with output vetting, security tiers, and regional policy. Scalattice routes, verifies, and bills per token for chat and per image for generation models. GPU providers host inference-only catalog models and earn the majority of developer spend on completed jobs.
The marketing site canonical URL is https://scalattice.com/. The same site is also served at https://scalattice.ai/, https://scalattice.app/, and https://scalattice.network/.
## Value proposition
### For API users (developers)
- OpenAI SDK drop-in: swap base URL and API key
- Requirements on chat completions: JSON, schema, contains, regex; retry then 422 ([blog](https://scalattice.com/blog/requirements/))
- Built-in output vetting, Scalattice runs multiple job replicas before responses return
- Security tiers (`tier1`, `tier2.5`) set per request or in the dashboard
- Regional policy headers: EU, US, APAC, or auto
- Published per-token catalog pricing with prepaid credits
- Image generation on `POST /v1/images/generations` (per-image billing, `b64_json`)
- Image edits on `POST /v1/images/edits` (reference images + prompt → pictures)
- Promotional offers: model-specific credits and unlimited access on select models ([Scalattice Cloud offers](https://scalattice.cloud/token/offers))
- Developer grants: apply to open programs; reviewed before credits ([Cloud grants](https://scalattice.cloud/token/grants), [how grants work](https://scalattice.com/blog/developer-grants/))
- Managed routing assigns provider capacity for each request
- Capacity fallback when provider GPUs are short
### For GPU providers
- Inference-only catalog models on your hardware
- The majority of developer token spend on completed inference jobs
- $0 connection fee
- Curated model catalog with demand ratings (1 to 5)
- Model weights mirrored by Scalattice
- Schedule control: always, paused, or UTC hour windows
- Open-source agent (MIT) for Linux, Windows, and macOS Apple Silicon (Intel Macs are not supported): https://github.com/scalattice/scalattice-agent
## Main
- [Home](https://scalattice.com/): Product overview, model catalog, and how the network works
- [Developers](https://scalattice.com/developers/): OpenAI-compatible API, vetting, security tiers, and regional policy
- [CLI](https://scalattice.com/cli/): Developer CLI install (npm + curl) and GitHub source
- [DHS / Don't Hit Send](https://scalattice.com/dont-hit-send/): Type and the model answers as you type. Unsigned visitors get a few ambiguous words, then a prompt to continue on Scalattice Cloud as a developer for live billed inference.
- [Providers](https://scalattice.com/providers/): Inference-only GPU hosting, majority revenue share, and $0 connection fee
- [Models](https://scalattice.com/models/): Full catalog with specs, strengths, live rates, and rival comparisons
- [Pricing](https://scalattice.com/pricing/): Published per-token rates for developers and provider economics
- [Blog](https://scalattice.com/blog/): Guides on models, inference, and hosting inference on your GPU
- [Network](https://scalattice.com/network/): Regional routing, vetting, security policy, and capacity fallback
- [Safety](https://scalattice.com/safety/): Host safety FAQ. Inference-only catalog jobs, tier2.5 split inference (lower prefill / upper decode), and what they do not
- [Company](https://scalattice.com/company/): Mission and how developers and providers connect through Scalattice
- [Support](https://scalattice.com/support/): Help centre and contact
- [Status](https://scalattice.com/health/): Live Cloud health. Same probe as scalattice.cloud/health, hosted here if Cloud is down
## Blog
- [Developer grants](https://scalattice.com/blog/developer-grants/): Apply to open programs; credits awarded after review
- [Requirements](https://scalattice.com/blog/requirements/): Fail-closed checks on chat completions (`json`, schema, regex)
- [Image models](https://scalattice.com/blog/image-models/): Image generation on Scalattice, starting with Qwen Image 2512 (`POST /v1/images/generations`, per-image billing)
- [Modelmeter](https://scalattice.com/blog/modelmeter/): Open-sourced lab for comparing open-model prices and latency
- [August 2026 catalog](https://scalattice.com/blog/catalog-aug-2026/): Why the current families are on the list, plus snapshot rates
- [LiteLLM + Open WebUI](https://scalattice.com/blog/litellm-open-webui/): Connect Open WebUI or LiteLLM to `https://api.scalattice.cloud/v1` with an `slt_` key
- [CLI + MCP](https://scalattice.com/blog/cli-mcp/): Terminal setup, npm install, optional MCP for Cursor / Claude Desktop
- [OpenAI SDK → Scalattice](https://scalattice.com/blog/openai-sdk-scalattice/): Drop-in base URL + API key setup (Python + curl)
- [Choosing a model](https://scalattice.com/blog/choosing-a-model/): Catalog guide for Llama, Qwen3, Gemma 4, DeepSeek R1, gpt-oss, and Coder
- [What is inference](https://scalattice.com/blog/what-is-inference/): Tokens, context, streaming, and pricing
- [Host inference on your GPU](https://scalattice.com/blog/host-inference-on-your-gpu/): Agent install, schedules, and earning in the pool
## Cloud
- [Scalattice Cloud](https://scalattice.cloud/): API keys, security policy defaults, credits, usage, provider earnings, and agent onboarding
- [Cloud status](https://scalattice.cloud/health): Same live probe; the public replica is [scalattice.com/health](https://scalattice.com/health/)
- [Cloud offers](https://scalattice.cloud/token/offers): Live promos and model-specific credit grants
- [Developer grants](https://scalattice.cloud/token/grants): Apply to open programs; answers are reviewed before credits are awarded
- [How grants work](https://scalattice.com/blog/developer-grants/): What grants are, who they are for, and how to apply
- [Developer docs](https://scalattice.cloud/docs/developers): API reference, Open WebUI, LiteLLM, routing headers, billing, host safety
- [Host safety](https://scalattice.cloud/docs/developers#host-safety): What runs on the agent; tier1 vs tier2.5
- [Open WebUI docs](https://scalattice.cloud/docs/developers#open-webui): Admin → Connections → `https://api.scalattice.cloud/v1`
- [LiteLLM docs](https://scalattice.cloud/docs/developers#litellm): `openai/` model prefix + Scalattice `api_base`
- [Provider docs](https://scalattice.cloud/docs/providers): Agent install, models, schedules, earnings
- [Provider host safety](https://scalattice.cloud/docs/providers#host-safety): Inference-only catalog jobs; no arbitrary customer containers
## Open source
- [Scalattice Agent](https://github.com/scalattice/scalattice-agent): GPU provider agent (Rust, MIT)
- [Scalattice CLI](https://github.com/scalattice/scalattice-cli): Interactive `scalattice>` prompt (`setup`, `whoami`, `credits`), optional MCP
- [npm: scalattice-cli](https://www.npmjs.com/package/scalattice-cli): Install with `npm install -g scalattice-cli`
- [CLI landing](https://scalattice.com/cli/): Install options and links
- [Cloud installer](https://scalattice.cloud/install/cli): `curl -fsSL … | sh`
## Agent onboarding (CLI)
```bash
curl -fsSL https://scalattice.cloud/install/cli | sh
# or: npm install -g scalattice-cli
scalattice
# prompt: setup, whoami, credits, help, exit
eval "$(scalattice init)"
```
Then call OpenAI-compatible `POST /v1/chat/completions` or `GET /v1/credits`. Optional: `scalattice mcp` for Claude Desktop / Cursor tools (MCP does not create accounts).
## Policies
- [Safety](https://scalattice.com/safety/): Host safety FAQ. Inference-only jobs and tier2.5 split inference
- [Privacy](https://scalattice.com/privacy/): Personal data and cookies
- [Terms](https://scalattice.com/terms/): Terms and conditions
- [Legal](https://scalattice.com/legal/): Legal information
- [Security](https://scalattice.com/security/): Vulnerability disclosure policy
- [Accessibility](https://scalattice.com/accessibility/): WCAG commitment and contact
## Optional
- [Sitemap](https://scalattice.com/sitemap.xml): Full list of public pages
- [Robots](https://scalattice.com/robots.txt): Crawler rules