LLaMA
An open-source large language model series released by Meta with publicly available model weights. It serves as one of the foundational components of Meta's AI technology stack.
Core Facts
Meta has open-sourced the LLaMA series of models.
The GPT-4 series uses the cl100k_base vocabulary (approximately 100,000 tokens), while the GPT-4o series upgrades to o200k_base (approximately 200,000 tokens).
Llama was released by Meta, and from the Llama 1 to the Llama 3 series, parameter scales range from 7B to 405B
Timeline (last 90 days)
The arXiv number for the LLaMA paper published by Meta is 2302.13971.
The weights of Meta's LLaMA series models are accessible to researchers, but come with commercial use restrictions and do not meet OSI's full open-source definition.
The advantage of local model deployment is that data never leaves the device and no subscription fees are required, but the hardware requirements are high — typically at least 8 GB of VRAM or a high-performance mobile chip.
RoPE has been widely adopted by mainstream models such as LLaMA, Qwen, and Mistral, and has given rise to length extrapolation variants such as YaRN and LongRoPE.
Local deployment typically refers to running open-source large models on the user's own device or private server, with typical options including the Ollama framework paired with models such as LLaMA and Mistral.
Locally deployed open-source models (such as the Llama series) can achieve truly local private conversations where data never leaves the local machine.
Meta's previously launched LLaMA series of large language models attracted a large number of researchers and developers through open weights.
The proliferation of open-source models such as Meta's LLaMA series and Mistral has lowered the technical barrier to launching distillation attacks by providing ready-made student model bases.
The paper uses Two One-Sided Tests (TOST) to elevate the indistinguishability finding to an equivalence claim, which holds across all 12 Qwen/LLaMA experimental cells.
OpenAI does not manufacture end-user hardware, models themselves are increasingly becoming commoditized, and open-source alternatives such as Meta's LLaMA series continue to lower switching costs
40 more timeline events
All Facts (20)
Meta has open-sourced the LLaMA series of models.
90%VerifiedThe GPT-4 series uses the cl100k_base vocabulary (approximately 100,000 tokens), while the GPT-4o series upgrades to o200k_base (approximately 200,000 tokens).
90%VerifiedLlama was released by Meta, and from the Llama 1 to the Llama 3 series, parameter scales range from 7B to 405B
90%VerifiedMeta's Llama series has been open-sourced since 2023, with the Llama 3.1 version reaching a maximum parameter count of 405B.
85%VerifiedFor applications with high call volumes, the total cost of self-hosted open-source model deployment is often significantly lower than using commercial APIs.
80%Verifiedllama.cpp是由开发者Georgi Gerganov于2023年发起的开源项目，使用纯C/C++实现了LLaMA系列模型的推理
80%VerifiedMeta的Llama系列使用自定义的Llama Community License，对月活超过7亿的企业有额外限制
80%Verified当前所有主流大模型（GPT、BERT、LLaMA等）都基于Transformer架构
80%VerifiedMeta's Llama series, Mistral, and several mainstream domestic model teams have all adopted an open-weight strategy.
80%VerifiedOpen-weight models refer to models whose parameters and weights are publicly released and allow users to download and deploy them independently, such as Meta's Llama series and Mistral.
75%VerifiedThe Llama series is a foundational model open-sourced by Meta.
75%VerifiedRepresentative open-source AI models include Meta's Llama series, Mistral, and Stability AI.
75%VerifiedWhen Meta previously released LLaMA 2 and LLaMA 3, it used a custom community license that set additional authorization thresholds for companies with more than 700 million monthly active users.
75%VerifiedMeta open-sourced the Llama series of models, Alibaba open-sourced the Qwen series, and Mistral AI has also adopted an open-source strategy.
75%VerifiedRoPE has been widely adopted by mainstream models including GPT-NeoX, LLaMA, and DeepSeek.
75%VerifiedMeta's FAIR lab is the birthplace of the open-source large model LLaMA series.
75%VerifiedN8N natively supports LangChain integration, has built-in AI Agent nodes, and can directly connect to mainstream models such as OpenAI and Anthropic
70%VerifiedMainstream open-source models such as Meta's Llama series, DeepSeek-R1, and Mistral mostly fall under the category of open-weight models.
70%VerifiedThe core technical foundation of generative AI is large language models, such as OpenAI's GPT series, Google's Gemini, and Meta's LLaMA.
70%VerifiedMeta released the LLaMA series of models in 2023, after which Mistral, Qwen, DeepSeek, and others followed in succession to form an open-source ecosystem.
70%