AI Radar

Llama 3.1

Available · Language

Llama 3.1 is a collection of Meta language models in 8B, 70B and 405B sizes, released on 23 July 2024 as open weights under the custom Llama 3.1 Community License and through more than 25 partners. It upgraded the Meta Llama 3 8B and 70B models, added a 405B model and raised the context length to 128K tokens, with support for eight languages. [1] [2] [3] Primary source

Timeline of Llama 3.1 →

Claims and evidence

  • Released Primary source[1]page date; Introducing Llama 3.1 [2]Model Release Date [4]README, Llama Models table
  • Status AvailableMeta publishes no deprecation page for Llama models; the weights of all three sizes are still listed in Meta's Hugging Face organisation (gated, license acceptance required) as seen on 2026-10-01. Primary source[3]
  • Successor of Meta Llama 3 Primary source[1]Introducing Llama 3.1 (upgraded versions of the 8B and 70B models) [5]Upgrading your application from Llama 3 to Llama 3.1
  • Change · Context length Context length increased from 8K to 128K tokens.Compared with Meta Llama 3 Primary source[4]README, Llama Models table [1]Takeaways
  • Change · Languages Official support for eight languages; the Meta Llama 3 model card named English as the intended language.Compared with Meta Llama 3 Primary source[2]
  • Change · Size Adds a 405B-parameter model next to upgraded 8B and 70B models.Compared with Meta Llama 3 Primary source[1]Introducing Llama 3.1
  • Change · Tool use Meta describes the upgraded 8B and 70B models as having tool use, and its documentation names function calling as a new Llama 3.1 feature.Compared with Meta Llama 3 Primary source[1]Introducing Llama 3.1 [5]Upgrading your application from Llama 3 to Llama 3.1
  • Input text Primary source[2]Model Information table, Input modalities
  • Output text, code Primary source[2]Model Information table, Output modalities
  • Open weights YesWeights under the custom Llama 3.1 Community License, after accepting the license. Primary source[1]Introducing Llama 3.1; Openness drives innovation [3]
  • Context window 128K tokens Primary source[1]Takeaways; Introducing Llama 3.1 [2]Model Information table, Context length (128k)
  • Access Open-weights download, cloud partner, consumer appConsumer access: Meta invited users in the US to try Llama 3.1 405B on WhatsApp and meta.ai. Primary source[1]Takeaways; Introducing Llama 3.1

Lineage

Predecessors

Successors

Based on

Not derived from another model.

Variants and derived

Siblings

None recorded.

All ancestors

All descendants

Variants

Llama 3.1 8B

Same dates as Llama 3.1.

  • Variant Inline variant in this record.Pretrained and instruction-tuned versions; Hugging Face repositories first named Meta-Llama-3.1-8B and Meta-Llama-3.1-8B-Instruct. Primary source[1] [2]Model Information [3]

Also known as: Llama-3.1-8B, Llama-3.1-8B-Instruct, Meta-Llama-3.1-8B, Meta-Llama-3.1-8B-Instruct

Differs in:

  • Parameters: 8B Evidence not assessed [2]Model Information table, Params

Llama 3.1 70B

Same dates as Llama 3.1.

  • Variant Inline variant in this record.Pretrained and instruction-tuned versions; Hugging Face repositories first named Meta-Llama-3.1-70B and Meta-Llama-3.1-70B-Instruct. Primary source[1] [2]Model Information [3]

Also known as: Llama-3.1-70B, Llama-3.1-70B-Instruct, Meta-Llama-3.1-70B, Meta-Llama-3.1-70B-Instruct

Differs in:

  • Parameters: 70B Evidence not assessed [2]Model Information table, Params

Llama 3.1 405B

Same dates as Llama 3.1.

  • Variant Inline variant in this record.Pretrained and instruction-tuned versions, each also as an FP8-quantized build, which Meta made for single-server-node inference. Meta's paper calls this model Llama 3 405B. Primary source[1] [2]Model Information [6]Abstract [3]

Also known as: Meta Llama 3.1 405B, Llama-3.1-405B, Llama-3.1-405B-Instruct, Llama-3.1-405B-FP8, Llama-3.1-405B-Instruct-FP8, Meta-Llama-3.1-405B, Meta-Llama-3.1-405B-Instruct, Meta-Llama-3.1-405B-Instruct-FP8, Llama 3 405B

Differs in:

  • Parameters: 405B Evidence not assessed [2]Model Information table, Params

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from Meta on AI Radar →