Meta-Llama-3-8B is the 8B parameter big language model of Meta Llama 3 series released by Meta AI on April 18, 2024. It is good at such complex tasks as language nuances, context understanding, code generation, translation and dialogue generation. Meta-Llama-3-8B-Instruct is the command tuning version of the 8B parameter.

Meta-Llama-3-8B is the 8B parameter big language model of Meta Llama 3 series released by Meta AI on April 18, 2024. It is good at such complex tasks as language nuances, context understanding, code generation, translation and dialogue generation. Meta-Llama-3-8B-Instruct is the command tuning version of the 8B parameter.

Meta-Llama-3-70B is the 70B parameter big language model of Meta Llama 3 series released by Meta AI on April 18, 2024. It is good at such complex tasks as language nuances, context understanding, code generation, translation and dialogue generation. Meta-Llama-3-70B-Instrument is the instruction tuning version of the 70B parameter.

Meta-Llama-3-70B is the 70B parameter big language model of Meta Llama 3 series released by Meta AI on April 18, 2024. It is good at such complex tasks as language nuances, context understanding, code generation, translation and dialogue generation. Meta-Llama-3-70B-Instrument is the instruction tuning version of the 70B parameter.

Compared with ERNIE 3.5, the flagship super large scale model developed by Baidu has achieved a comprehensive upgrade in model capability, which is widely applicable to complex task scenarios in various fields; Support automatic docking with Baidu search plug-in to ensure the timeliness of question and answer information.

Compared with ERNIE 3.5, the flagship super large scale model developed by Baidu has achieved a comprehensive upgrade in model capability, which is widely applicable to complex task scenarios in various fields; Support automatic docking with Baidu search plug-in to ensure the timeliness of question and answer information.

The flagship large-scale ⼤⾔⾔⾔⾔ model developed by Baidu itself covers a large number of Chinese and English language materials, and has a strong universal ability, which can meet the requirements of most dialogue questions and answers, creation generation, and plug-in application scenarios; Support automatic docking with Baidu search plug-in to ensure the timeliness of question and answer information.

The flagship large-scale ⼤⾔⾔⾔⾔ model developed by Baidu itself covers a large number of Chinese and English language materials, and has a strong universal ability, which can meet the requirements of most dialogue questions and answers, creation generation, and plug-in application scenarios; Support automatic docking with Baidu search plug-in to ensure the timeliness of question and answer information.

Baidu's self-developed high-performance big language model, which was newly released in 2024, has excellent general ability and is suitable for fine tuning as a base model to better handle specific scenario problems, and has excellent reasoning performance.

Baidu's self-developed high-performance big language model, which was newly released in 2024, has excellent general ability and is suitable for fine tuning as a base model to better handle specific scenario problems, and has excellent reasoning performance.

The lightweight big language model developed by Baidu gives consideration to excellent model effect and reasoning performance, and is suitable for reasoning with low computing power AI accelerator cards.

The lightweight big language model developed by Baidu gives consideration to excellent model effect and reasoning performance, and is suitable for reasoning with low computing power AI accelerator cards.

The ultra-high performance big language model developed by Baidu has the lowest deployment and fine tuning cost among Wenxin series models.

The ultra-high performance big language model developed by Baidu has the lowest deployment and fine tuning cost among Wenxin series models.

The vertical scene big language model developed by Baidu is suitable for application scenarios such as game NPC, customer service dialogue, dialogue role play, etc. It has a more distinctive and consistent style, stronger ability to follow instructions, and better reasoning performance.

The vertical scene big language model developed by Baidu is suitable for application scenarios such as game NPC, customer service dialogue, dialogue role play, etc. It has a more distinctive and consistent style, stronger ability to follow instructions, and better reasoning performance.

The vertical scene big language model developed by Baidu is suitable for the use of external tools and business function call scenarios in dialogue Q&A, with stronger synthetic ability of structured answers, more stable output format, and better reasoning performance.

The vertical scene big language model developed by Baidu is suitable for the use of external tools and business function call scenarios in dialogue Q&A, with stronger synthetic ability of structured answers, more stable output format, and better reasoning performance.

The special version model of Qianfan AppBuilder has carried out special command tuning for enterprise level large model applications, which can achieve better results under the same scale model in question and answer scenarios and agent related scenarios. It needs to be used together with the "Baidu Intelligent Cloud Qianfan AppBuilder" product or in combination with "AppBuilder SDK" alone

The special version model of Qianfan AppBuilder has carried out special command tuning for enterprise level large model applications, which can achieve better results under the same scale model in question and answer scenarios and agent related scenarios. It needs to be used together with the "Baidu Intelligent Cloud Qianfan AppBuilder" product or in combination with "AppBuilder SDK" alone

The cross modal image generation model independently developed by Baidu and the hybrid expert modeling for innovative knowledge enhancement are the world's first AI painting model for knowledge enhancement. It has significant advantages in semantic control, image details, Chinese understanding and other aspects, and has been used as a basic model in Wenxin Yige and other related businesses.

The cross modal image generation model independently developed by Baidu and the hybrid expert modeling for innovative knowledge enhancement are the world's first AI painting model for knowledge enhancement. It has significant advantages in semantic control, image details, Chinese understanding and other aspects, and has been used as a basic model in Wenxin Yige and other related businesses.

Embedding V1 is a text representation model based on Baidu Wenxin's big model technology, which converts text into vector form represented by numerical values, and is used in text retrieval, information recommendation, knowledge mining and other scenarios.

Embedding V1 is a text representation model based on Baidu Wenxin's big model technology, which converts text into vector form represented by numerical values, and is used in text retrieval, information recommendation, knowledge mining and other scenarios.

The compressed version of Qianfan team based on BLOOMZ-7B, integrating quantization, sparsity and other technologies, reduces the video memory occupation by more than 30%

The compressed version of Qianfan team based on BLOOMZ-7B, integrating quantization, sparsity and other technologies, reduces the video memory occupation by more than 30%

Qianfan team's Chinese enhanced version based on Llama-2-7b has excellent performance in CMMLU, C-EVAL and other Chinese data sets

Qianfan team's Chinese enhanced version based on Llama-2-7b has excellent performance in CMMLU, C-EVAL and other Chinese data sets

Qianfan team's enhanced version based on Qianfan Chinese Llama-2-7B supports 32K context

Qianfan team's enhanced version based on Qianfan Chinese Llama-2-7B supports 32K context

The 13 billion parameter pre training big language model developed by the Qianfan team based on the Llama2 architecture starts from scratch on the high-quality self owned corpus of 1T+tokens, supports bilingual Chinese and English, and the length of the context window is 4K. The overall reasoning effect is better than the open source model with the same parameter level

The 13 billion parameter pre training big language model developed by the Qianfan team based on the Llama2 architecture starts from scratch on the high-quality self owned corpus of 1T+tokens, supports bilingual Chinese and English, and the length of the context window is 4K. The overall reasoning effect is better than the open source model with the same parameter level

Qianfan team's Chinese enhanced version based on Llama-2-13b has excellent performance in CMMLU, C-EVAL and other Chinese data sets

Qianfan team's Chinese enhanced version based on Llama-2-13b has excellent performance in CMMLU, C-EVAL and other Chinese data sets

Qianfan team's Chinese enhanced version based on Llama-2-70b has excellent performance in CMMLU, C-EVAL and other Chinese data sets

Qianfan team's Chinese enhanced version based on Llama-2-70b has excellent performance in CMMLU, C-EVAL and other Chinese data sets

Qianfan Llama2-70B compressed is the compressed version of Qianfan team based on Llama2-70B, which integrates compression acceleration technologies such as quantification, sparsity, operator optimization, and greatly reduces the consumption of model reasoning resources.

Qianfan Llama2-70B compressed is the compressed version of Qianfan team based on Llama2-70B, which integrates compression acceleration technologies such as quantification, sparsity, operator optimization, and greatly reduces the consumption of model reasoning resources.