兔子API 兔子API
中

model

Meta Llama

Direct answerLlama is Meta's open-weight family, normally used through self-hosting or cloud partners; model cards, licenses, prompt templates, and the actual runtime jointly define behavior.

Updated · Reviewed

Family position

Llama is released as open weights for download, self-hosting, fine-tuning, or hosting through channels such as AWS and Azure. Meta does not provide one universal public REST API for every Llama release, so this page does not invent one.

Versions and modalities

Llama generations, parameter sizes, and vision variants have different tokenizers, context, prompt templates, hardware needs, and licenses. Community quantizations, fine-tunes, and distillations are not identical to Meta's original releases even when names look similar.

Choosing within the family

Private networks, offline inference, controlled deployment, and custom training are key advantages; teams without inference operations should evaluate a managed host. Measure quality, memory, time to first token, throughput, quantization loss, and safety.

API use

A Llama ID on this site uses the route defined by its channel adapter, commonly OpenAI-compatible chat. Self-hosting uses a runtime such as vLLM, TGI, or Ollama. Retain the weight release, digest, prompt template, and runtime parameters.

License and responsibility

Read the exact Community License, Acceptable Use Policy, and model card. The operating team owns access control, content safety, dependency vulnerabilities, provenance, monitoring, and upgrades.

Use cases

  • Open weights
  • Multimodality
  • Private deployment

API protocols

  • /v1/chat/completions

FAQ

Is the model name permanent?

No. Official aliases can move and enabled API IDs can change; pin a tested ID in production and verify availability separately in the model marketplace.

Does Llama have one official Meta API and a fixed request protocol?

There is no single public REST API covering every release. Self-hosted protocol comes from runtimes such as vLLM, TGI, or Ollama, while a cloud host defines its own protocol. Retain the weight version, digest, tokenizer, prompt template, and quantization settings.

Can every downloaded Llama or community fine-tune be used commercially?

Do not assume so. Read the exact weight's Community License, Acceptable Use Policy, and model card. Community quantizations, distillations, or fine-tunes can have different provenance and terms, and the deployer owns safety, access control, and operations.

  • Meta LlamaMeta's open-weight model family, available for self-hosting or through partner cloud platforms.
  • Choosing and calling coding modelsSelect coding models by patch correctness, test pass rate, tool success, latency, and cost on real repositories, not marketplace copy or a single benchmark.
  • Tool use and AI agent model guideAn agent is a controlled loop, not merely tool calling: structured schemas, least privilege, state, idempotency, timeouts, approval, audit, and replayable evaluation.

Official sources

  1. Meta Llama Official