# ALLM

> ALLM is an AI model intelligence API and public registry. It normalizes deployment-level pricing, capabilities, lifecycle, limits, evidence, and provider compatibility so teams can make explainable model decisions.

ALLM does not proxy inference traffic or act as an LLM gateway. Applications continue to send requests to their chosen provider or gateway; ALLM supplies the decision data and change intelligence around those requests.

## Public registry

- [Model registry](https://use-allm.com/registry): compare normalized deployments.
- [Canonical models](https://use-allm.com/models): model releases, provider availability, prices, capabilities, and lifecycle.
- [Providers](https://use-allm.com/providers): provider-specific deployment catalogs.
- [AI labs](https://use-allm.com/labs): model makers and official naming conventions.
- [Cost control beyond tokens](https://use-allm.com/cost-control): why voice sessions (billed per second) and agent sandboxes (billed per container) are invisible to token-only budgets, and which meters decide what an agent actually costs.
- [Intelligence reports](https://use-allm.com/intelligence): live, source-linked reports on deployment observations, lifecycle dates, capabilities, and provider differences.
  - [GPT-5.6 Sol: one model, six meters, and a price that expires](https://use-allm.com/intelligence/gpt-5-6-sol-price-deck): the published Global meter deck ($2 to $16 input, $10 to $60 output per million tokens), the documented 10% data-residency uplift, the "at least through November 21, 2026" promotional rate, and the $5 / $30 alias records the registry still carries.
  - [Voice sessions and agent sandboxes: the non-token meter](https://use-allm.com/intelligence/voice-sessions-and-agent-sandboxes): GPT-Live 1 voice sessions billed per minute and Agents API runs billed per container — the routes and meters a token-only gateway does not cover.
  - [September 2026: the three-day frontier wave](https://use-allm.com/intelligence/september-2026-frontier-wave): Claude Fable 5.1, Gemini 3.8 Flash and GPT-6 Astra shipped within 72 hours — launch contracts, pricing and registry coverage.
  - [Azure AI: the October 2026 retirement wave](https://use-allm.com/intelligence/azure-october-2026-retirement-wave): which Azure model versions actually retire in October 2026, and where the registry disagrees with Microsoft's schedule.
  - [This week in model deployments](https://use-allm.com/intelligence/weekly-deployment-watch): provider endpoints observed in the last seven days.
  - [Models with a published end-of-life date](https://use-allm.com/intelligence/published-model-retirements): concrete deployment retirement dates.
  - [Which endpoints support structured output?](https://use-allm.com/intelligence/structured-output-endpoints): deployment-level structured-output contracts.
  - [OpenAI vs Azure AI: the deployment delta](https://use-allm.com/intelligence/openai-vs-azure-ai): one canonical model across two provider surfaces.

## Documentation

- [ALLM documentation](https://docs.use-allm.com/): API concepts and implementation guides.
- [Quickstart](https://docs.use-allm.com/quickstart): make a first model-intelligence query.
- [Deployment model](https://docs.use-allm.com/concepts/deployments): why model releases and provider deployments differ.
- [Lifecycle and EOL](https://docs.use-allm.com/concepts/lifecycle): deprecation, retirement, coverage, and evidence.
- [Production use cases](https://docs.use-allm.com/use-cases): identity, safe failover, constraint routing, cost controls, and release drift.
- [OpenAPI specification](https://api.use-allm.com/openapi.json): machine-readable ALLM API contract.

## Primary concepts

- A model release identifies what a lab created.
- A deployment identifies a provider-specific model ID, region, API surface, rate card, capabilities, limits, lifecycle, and evidence.
- Pricing, capability, and lifecycle assertions are tied to the deployment, not assumed from a model name.
- Catalog releases are immutable snapshots intended for reproducible decisions and audit trails.

## Crawling

- [Full knowledge base](https://use-allm.com/llms-full.txt): the complete editorial body of ALLM in one Markdown file.
- [Sitemap](https://use-allm.com/sitemap.xml)
- [Robots policy](https://use-allm.com/robots.txt)