Open Source AI Models Timeline

From GPT-2 to GPT-OSS: Witness the milestone moments of open-source AI models

GPT-OSS-120B
Latest
2025-08
120B
OpenAI
OpenAI's return to open source, highly symbolic, excellent performance/efficiency
GitHubHuggingFace
Qwen3-235B
2025-04
235B
Alibaba Cloud
One of the strongest Chinese models, multimodal + large parameters, fully open source
GitHubHuggingFace
GLM-4-9B
2024-06
9B
Tsinghua University (Zhipu AI)
Tool calling capability close to GPT-4, excellent Chinese task performance
GitHubHuggingFace
DBRX
2024-03
132B (36B active)
Databricks / MosaicML
One of the strongest open-source MoE models, approaching GPT-4 performance
GitHubHuggingFace
Qwen1.5-72B
2024-03
72B
Alibaba Cloud
Extremely strong Chinese performance, supports long context, function calling
GitHubHuggingFace
Mixtral-8x7B
2023-12
45B active
Mistral AI
MoE sparse expert architecture, excellent inference speed/quality
GitHubHuggingFace
Mistral-7B
2023-09
7B
Mistral AI
Excellent performance, architectural improvements, industry-leading parameter efficiency
GitHubHuggingFace
LLaMA 2 70B
2023-07
70B
Meta
Truly promoted open-source large model deployment, most widely used by community
GitHubHuggingFace
YaLM-100B
2022-06
100B
Yandex
Open-source Russian/English GPT model, industrial scale
GitHubHuggingFace
BLOOM
2022-07
176B
BigScience (HuggingFace)
Multilingual, community-driven open-source large model, exemplar of open collaboration
GitHubHuggingFace
OPT-175B
2022-05
175B
Meta
Model + data + training logs all open, high transparency
GitHub
GPT-NeoX-20B
2021-12
20B
EleutherAI
Large-scale open-source GPT model with high text generation quality
GitHubHuggingFace
GPT-J-6B
2021-06
6B
EleutherAI
First open-source model with performance close to GPT-3
GitHubHuggingFace
GPT-2
2019-02
1.5B
OpenAI
First widely influential open-weight Transformer model, laid the foundation
GitHubHuggingFace
Slide to view more models