BotLineage
Search…
Guide · 9 models

Best open source LLMs

Open-weight models publish their trained weights, so you can download, self-host, fine-tune and run them without sending data to a third party. Licences vary — some are fully permissive, others restrict commercial use at scale — so check each model's terms.

The list below is drawn straight from the BotLineage database and ordered newest first. Newer releases generally beat older ones of the same size, and context windows have grown from a few thousand tokens to millions.

  1. 01

    Llama 4

    Apr 5, 2025 · 10M context · Open weights

    Meta's Llama 4 MoE family with very long context and multimodal input.

  2. 02

    DeepSeek R1

    Jan 20, 2025 · 128K context · Open weights

    Open-weight reasoning model that popularized reinforcement-learned chain-of-thought at scale.

  3. 03

    DeepSeek V3

    Dec 26, 2024 · 128K context · Open weights

    671B MoE open-weight base model.

  4. 04

    Llama 3

    Apr 18, 2024 · 8.2K context · Open weights

    8B and 70B open-weight release, later extended to Llama 3.1 and 3.2 with vision.

  5. 05

    Mixtral 8x7B

    Dec 11, 2023 · 32K context · Open weights

    Sparse mixture-of-experts open-weight model from Mistral.

  6. 06

    Grok-1

    Nov 4, 2023 · 8.2K context · Open weights

    First Grok model from xAI, later open-sourced.

  7. 07

    Mistral 7B

    Sep 27, 2023 · 8.2K context · Open weights

    Compact 7B open-weight model that punched above its parameter count.

  8. 08

    Llama 2

    Jul 18, 2023 · 4.1K context · Open weights

    First widely-licensed open-weight Llama release.

  9. 09

    GPT-2

    Feb 14, 2019 · 1.0K context · Open weights

    Second-generation GPT model, initially released with a staged rollout over safety concerns.

See also: LLM context window comparison · All comparisons