Meta

The Llama family

Meta's Llama models, which helped popularize open weights; since 2026 Meta has also released models under the Muse name.

Only versions we have a record for. Missing ones appear as records are added.

12versions2023–2026Latest · Muse Spark Open

At a glance

Where labels would overlap, only the dot is drawn — point at it to see the name; every version is in the list below.

12 versions

In release order. Each date comes from the record that documents it — open a name to see the record.

  1. LLaMA 1

    Shared with vetted researchers for non-commercial use

    LlamaText
  2. Llama 2

    Commercial use allowed, under a Community License

    LlamaOpen weights
  3. Code Llama

    A code model built on Llama 2

    LlamaCodeOpen weights
  4. Llama 3

    8B and 70B sizes; raised the bar for local use

    LlamaOpen weights
  5. Llama 3.1

    Up to 405B; license allowed training on its outputs

    LlamaOpen weights
  6. Llama 3.2

    Image-reading 11B and 90B, plus 1B and 3B for phones

    LlamaMultimodalOpen weights
  7. Llama 3.3

    A text-only 70B model

    LlamaTextOpen weights
  8. Llama 4 Scout·Maverick

    Released amid controversy over an unreleased LMArena entry

    LlamaOpen weights
  9. Muse Spark

    Meta's multimodal reasoning model; weights not released

    MuseMultimodalReasoning
  10. Muse Glimmer 30B

    First open weights since Llama 4, under Apache 2.0

    MuseOpen weightsMultimodal
  11. Muse Spark 1.3

    Update the company says uses fewer tool calls and tokens

    MuseReasoning
  12. Muse Spark Open

    Muse Spark released as open weights six months on (42B active)

    LatestMuseOpen weightsMultimodal