stair-lab/platinum_detect · Datasets at Hugging Face

name stringlengths 10 47	display_name stringlengths 5 41	short_display_name stringlengths 5 41 ⌀	description stringlengths 3 534	creator_organization stringlengths 4 22	access stringclasses 3 values	todo bool 2 classes	release_date stringdate 2019-10-23 00:00:00 2024-12-11 00:00:00 ⌀	num_parameters float64 1 1,210B ⌀
huggingface/santacoder	SantaCoder (1.1B)	null	SantaCoder (1.1B parameters) model trained on the Python, Java, and JavaScript subset of The Stack (v1.1) ([model card](https://huggingface.co/bigcode/santacoder)).	BigCode	open	false	null	null
huggingface/starcoder	StarCoder (15.5B)	null	The StarCoder (15.5B parameter) model trained on 80+ programming languages from The Stack (v1.2) ([model card](https://huggingface.co/bigcode/starcoder)).	BigCode	open	false	null	null
HuggingFaceM4/idefics-80b	IDEFICS (80B)	null	IDEFICS (80B parameters) is an open-source model based on DeepMind's Flamingo. ([blog](https://huggingface.co/blog/idefics))	HuggingFace	open	false	2023/8/22	80,000,000,000
HuggingFaceM4/idefics-80b-instruct	IDEFICS instruct (80B)	null	IDEFICS instruct (80B parameters) is an open-source model based on DeepMind's Flamingo. ([blog](https://huggingface.co/blog/idefics))	HuggingFace	open	false	2023/8/22	80,000,000,000
HuggingFaceM4/idefics-9b	IDEFICS (9B)	null	IDEFICS (9B parameters) is an open-source model based on DeepMind's Flamingo. ([blog](https://huggingface.co/blog/idefics))	HuggingFace	open	false	2023/8/22	9,000,000,000
HuggingFaceM4/idefics-9b-instruct	IDEFICS instruct (9B)	null	IDEFICS instruct (9B parameters) is an open-source model based on DeepMind's Flamingo. ([blog](https://huggingface.co/blog/idefics))	HuggingFace	open	false	2023/8/22	9,000,000,000
HuggingFaceM4/idefics2-8b	IDEFICS 2 (8B)	IDEFICS 2 (8B)	IDEFICS 2 (8B parameters) is an open multimodal model that accepts arbitrary sequences of image and text inputs and produces text outputs. ([blog](https://huggingface.co/blog/idefics2)).	HuggingFace	open	false	2024/4/15	8,000,000,000
lightningai/lit-gpt	Lit-GPT	null	Lit-GPT is an optimized collection of open-source LLMs for finetuning and inference. It supports – Falcon, Llama 2, Vicuna, LongChat, and other top-performing open-source large language models.	Lightning AI	open	false	2023/4/4	1
lmsys/vicuna-13b-v1.3	Vicuna v1.3 (13B)	null	Vicuna v1.3 (13B) is an open-source chatbot trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.	LMSYS	open	false	2023/6/22	13,000,000,000
lmsys/vicuna-13b-v1.5	Vicuna 13B (v1.5)	Vicuna 13B (v1.5)	Vicuna is a large language model trained on 13 billion parameters. ([blog](https://lmsys.com/vicuna-13b-v1.5/))	LMSYS	open	false	2024/4/18	13,000,000,000
lmsys/vicuna-7b-v1.3	Vicuna v1.3 (7B)	null	Vicuna v1.3 (7B) is an open-source chatbot trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.	LMSYS	open	false	2023/6/22	7,000,000,000
lmsys/vicuna-7b-v1.5	Vicuna 7B (v1.5)	Vicuna 7B (v1.5)	Vicuna is a large language model trained on 7 billion parameters. ([blog](https://lmsys.com/vicuna-7b-v1.5/))	LMSYS	open	false	2024/4/18	7,000,000,000
meta-llama/Llama-2-13b-chat-hf	Llama 2 13B Chat	Llama 2 13B Chat	Llama 2 13B Chat is a large language model trained on 13 billion parameters. ([blog](https://meta-llama.com/llama-2-13b-chat-hf/))	Meta Llama	open	false	2024/4/18	13,000,000,000
meta-llama/Llama-2-70b-chat-hf	Llama 2 70B Chat	Llama 2 70B Chat	Llama 2 70B Chat is a large language model trained on 70 billion parameters. ([blog](https://meta-llama.com/llama-2-70b-chat-hf/))	Meta Llama	open	false	2024/4/18	70,000,000,000
meta-llama/Llama-2-7b-chat-hf	Llama 2 7B Chat	Llama 2 7B Chat	Llama 2 7B Chat is a large language model trained on 7 billion parameters. ([blog](https://meta-llama.com/llama-2-7b-chat-hf/))	Meta Llama	open	false	2024/4/18	7,000,000,000
meta/llama-13b	LLaMA (13B)	null	LLaMA is a collection of foundation language models ranging from 7B to 65B parameters.	Meta	open	false	2023/2/24	13,000,000,000
meta/llama-2-13b	Llama 2 (13B)	null	Llama 2 pretrained models are trained on 2 trillion tokens, and have double the context length than Llama 1.	Meta	open	false	2023/7/18	13,000,000,000
meta/llama-2-70b	Llama 2 (70B)	null	Llama 2 pretrained models are trained on 2 trillion tokens, and have double the context length than Llama 1.	Meta	open	false	2023/7/18	70,000,000,000
meta/llama-2-7b	Llama 2 (7B)	null	Llama 2 pretrained models are trained on 2 trillion tokens, and have double the context length than Llama 1.	Meta	open	false	2023/7/18	7,000,000,000
meta/llama-3-70b	Llama 3 (70B)	Llama 3 (70B)	Llama 3 is a family of language models that have been trained on more than 15 trillion tokens, and use Grouped-Query Attention (GQA) for improved inference scalability. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)	Meta	open	false	2024/4/18	70,000,000,000
meta/llama-3-70b-chat	Llama 3 Instruct (70B)	Llama 3 Instruct (70B)	Llama 3 is a family of language models that have been trained on more than 15 trillion tokens, and use Grouped-Query Attention (GQA) for improved inference scalability. It used SFT, rejection sampling, PPO and DPO for post-training. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)	Meta	open	false	2024/4/18	70,000,000,000
meta/llama-3-8b	Llama 3 (8B)	Llama 3 (8B)	Llama 3 is a family of language models that have been trained on more than 15 trillion tokens, and use Grouped-Query Attention (GQA) for improved inference scalability. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)	Meta	open	false	2024/4/18	8,000,000,000
meta/llama-3-8b-chat	Llama 3 Instruct (8B)	Llama 3 Instruct (8B)	Llama 3 is a family of language models that have been trained on more than 15 trillion tokens, and use Grouped-Query Attention (GQA) for improved inference scalability. It used SFT, rejection sampling, PPO and DPO for post-training. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)	Meta	open	false	2024/4/18	8,000,000,000
meta/llama-3.1-405b-instruct-turbo	Llama 3.1 Instruct Turbo (405B)	Llama 3.1 Instruct Turbo (405B)	Llama 3.1 (405B) is part of the Llama 3 family of dense Transformer models that that natively support multilinguality, coding, reasoning, and tool usage. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/), [blog](https://ai.meta.com/blog/meta-llama-3-1/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))	Meta	open	false	2024/7/23	405,000,000,000
meta/llama-3.1-70b-instruct-turbo	Llama 3.1 Instruct Turbo (70B)	Llama 3.1 Instruct Turbo (70B)	Llama 3.1 (70B) is part of the Llama 3 family of dense Transformer models that that natively support multilinguality, coding, reasoning, and tool usage. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/), [blog](https://ai.meta.com/blog/meta-llama-3-1/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))	Meta	open	false	2024/7/23	70,000,000,000
meta/llama-3.1-8b-instruct-turbo	Llama 3.1 Instruct Turbo (8B)	Llama 3.1 Instruct Turbo (8B)	Llama 3.1 (8B) is part of the Llama 3 family of dense Transformer models that that natively support multilinguality, coding, reasoning, and tool usage. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/), [blog](https://ai.meta.com/blog/meta-llama-3-1/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))	Meta	open	false	2024/7/23	8,000,000,000
meta/llama-3.2-11b-vision-instruct-turbo	Llama 3.2 Vision Instruct Turbo (11B)	Llama 3.2 Vision Instruct Turbo (11B)	The Llama 3.2 Vision collection of multimodal large language models (LLMs) is a collection of pretrained and instruction-tuned image reasoning generative models in 11B and 90B sizes. ([blog](https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))	Meta	open	false	2024/9/25	10,700,000,000
meta/llama-3.2-90b-vision-instruct-turbo	Llama 3.2 Vision Instruct Turbo (90B)	Llama 3.2 Vision Instruct Turbo (90B)	The Llama 3.2 Vision collection of multimodal large language models (LLMs) is a collection of pretrained and instruction-tuned image reasoning generative models in 11B and 90B sizes. ([blog](https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))	Meta	open	false	2024/9/25	88,600,000,000
meta/llama-3.3-70b-instruct-turbo	Llama 3.3 Instruct Turbo (70B)	Llama 3.3 Instruct Turbo (70B)	Llama 3.3 (70B) is part of the Llama 3 family of dense Transformer models that that natively support multilinguality, coding, reasoning, and tool usage. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))	Meta	open	false	2024/12/6	70,000,000,000
meta/llama-30b	LLaMA (30B)	null	LLaMA is a collection of foundation language models ranging from 7B to 65B parameters.	Meta	open	false	2023/2/24	30,000,000,000
meta/llama-65b	LLaMA (65B)	null	LLaMA is a collection of foundation language models ranging from 7B to 65B parameters.	Meta	open	false	2023/2/24	65,000,000,000
meta/llama-7b	LLaMA (7B)	null	LLaMA is a collection of foundation language models ranging from 7B to 65B parameters.	Meta	open	false	2023/2/24	7,000,000,000
microsoft/llava-1.5-13b-hf	LLaVA 1.5 (13B)	LLaVA 1.5 (13B)	LLaVa is an open-source chatbot trained by fine-tuning LlamA/Vicuna on GPT-generated multimodal instruction-following data. ([paper](https://arxiv.org/abs/2304.08485))	Microsoft	open	false	2023/10/5	13,000,000,000
microsoft/phi-2	Phi-2	Phi-2	Phi-2 is a Transformer with 2.7 billion parameters. It was trained using the same data sources as Phi-1.5, augmented with a new data source that consists of various NLP synthetic texts and filtered websites (for safety and educational value)	Microsoft	open	false	2023/10/5	13,000,000,000
microsoft/phi-3-medium-4k-instruct	Phi-3 (14B)	Phi-3 (14B)	Phi-3-Medium-4K-Instruct is a lightweight model trained with synthetic data and filtered publicly available website data with a focus on high-quality and reasoning dense properties. ([paper](https://arxiv.org/abs/2404.14219), [blog](https://azure.microsoft.com/en-us/blog/new-models-added-to-the-phi-3-family-available-on-microsoft-azure/))	Microsoft	open	false	2024/5/21	14,000,000,000
microsoft/phi-3-small-8k-instruct	Phi-3 (7B)	Phi-3 (7B)	Phi-3-Small-8K-Instruct is a lightweight model trained with synthetic data and filtered publicly available website data with a focus on high-quality and reasoning dense properties. ([paper](https://arxiv.org/abs/2404.14219), [blog](https://azure.microsoft.com/en-us/blog/new-models-added-to-the-phi-3-family-available-on-microsoft-azure/))	Microsoft	open	false	2024/5/21	7,000,000,000
microsoft/TNLGv2_530B	TNLG v2 (530B)	null	TNLG v2 (530B parameters) autoregressive language model trained on a filtered subset of the Pile and CommonCrawl ([paper](https://arxiv.org/pdf/2201.11990.pdf)).	Microsoft/NVIDIA	closed	false	2022/1/28	530,000,000,000
microsoft/TNLGv2_7B	TNLG v2 (6.7B)	null	TNLG v2 (6.7B parameters) autoregressive language model trained on a filtered subset of the Pile and CommonCrawl ([paper](https://arxiv.org/pdf/2201.11990.pdf)).	Microsoft/NVIDIA	closed	false	2022/1/28	6,700,000,000
mistralai/mistral-7b-instruct-v0.1	Mistral Instruct v0.1 (7B)	Mistral Instruct v0.1 (7B)	Mistral v0.1 Instruct 7B is a 7.3B parameter transformer model that uses Grouped-Query Attention (GQA) and Sliding-Window Attention (SWA). The instruct version was fined-tuned using publicly available conversation datasets. ([blog post](https://mistral.ai/news/announcing-mistral-7b/))	Mistral AI	open	false	2023/9/27	7,300,000,000
mistralai/Mistral-7B-Instruct-v0.2	Mistral 7B Instruct v0.2	Mistral 7B Instruct v0.2	Mistral 7B Instruct v0.2 is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://mistral.ai/mistral-7b-instruct-v0-2/))	Mistral AI	open	false	2024/4/18	7,000,000,000
mistralai/mistral-7b-instruct-v0.3	Mistral Instruct v0.3 (7B)	Mistral Instruct v0.3 (7B)	Mistral v0.3 Instruct 7B is a 7.3B parameter transformer model that uses Grouped-Query Attention (GQA). Compared to v0.1, v0.2 has a 32k context window and no Sliding-Window Attention (SWA). ([blog post](https://mistral.ai/news/la-plateforme/))	Mistral AI	open	false	2024/5/22	7,300,000,000
mistralai/mistral-7b-v0.1	Mistral v0.1 (7B)	null	Mistral 7B is a 7.3B parameter transformer model that uses Grouped-Query Attention (GQA) and Sliding-Window Attention (SWA).	Mistral AI	open	false	2023/9/27	7,300,000,000
mistralai/Mistral-7B-v0.1	Mistral 7B v0.1	Mistral 7B v0.1	Mistral 7B v0.1 is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://mistral.ai/mistral-7b-v0-1/))	Mistral AI	open	false	2024/4/18	7,000,000,000
mistralai/mistral-large-2402	Mistral Large (2402)	Mistral Large (2402)	Mistral Large is a multilingual model with a 32K tokens context window and function-calling capabilities. ([blog](https://mistral.ai/news/mistral-large/))	Mistral AI	limited	false	2023/2/26	null
mistralai/mistral-large-2407	Mistral Large 2 (2407)	Mistral Large 2 (2407)	Mistral Large 2 is a 123 billion parameter model that has a 128k context window and supports dozens of languages and 80+ coding languages. ([blog](https://mistral.ai/news/mistral-large-2407/))	Mistral AI	open	false	2023/7/24	123,000,000,000
mistralai/mistral-medium-2312	Mistral Medium (2312)	Mistral Medium (2312)	Mistral is a transformer model that uses Grouped-Query Attention (GQA) and Sliding-Window Attention (SWA).	Mistral AI	limited	false	2023/12/11	null
mistralai/mistral-small-2402	Mistral Small (2402)	Mistral Small (2402)	Mistral Small is a multilingual model with a 32K tokens context window and function-calling capabilities. ([blog](https://mistral.ai/news/mistral-large/))	Mistral AI	limited	false	2023/2/26	null
mistralai/mixtral-8x22b	Mixtral (8x22B)	Mixtral (8x22B)	Mistral AI's mixture-of-experts model that uses 39B active parameters out of 141B ([blog post](https://mistral.ai/news/mixtral-8x22b/)).	Mistral AI	open	false	2024/4/10	176,000,000,000
mistralai/mixtral-8x22b-instruct-v0.1	Mixtral Instruct (8x22B)	Mixtral Instruct (8x22B)	Mistral AI's mixture-of-experts model that uses 39B active parameters out of 141B ([blog post](https://mistral.ai/news/mixtral-8x22b/)).	Mistral AI	open	false	2024/4/10	176,000,000,000
mistralai/mixtral-8x7b-32kseqlen	Mixtral (8x7B 32K seqlen)	Mixtral (8x7B 32K seqlen)	Mixtral is a mixture-of-experts model that has 46.7B total parameters but only uses 12.9B parameters per token. ([blog post](https://mistral.ai/news/mixtral-of-experts/), [tweet](https://twitter.com/MistralAI/status/1733150512395038967)).	Mistral AI	open	false	2023/12/8	46,700,000,000
mistralai/mixtral-8x7b-instruct-v0.1	Mixtral Instruct (8x7B)	Mixtral Instruct (8x7B)	Mixtral Instruct (8x7B) is a version of Mixtral (8x7B) that was optimized through supervised fine-tuning and direct preference optimisation (DPO) for careful instruction following. ([blog post](https://mistral.ai/news/mixtral-of-experts/)).	Mistral AI	open	false	2023/12/11	46,700,000,000
mistralai/open-mistral-nemo-2407	Mistral NeMo (2402)	Mistral NeMo (2402)	Mistral NeMo is a multilingual 12B model with a large context window of 128K tokens. ([blog](https://mistral.ai/news/mistral-nemo/))	Mistral AI	open	false	2024/7/18	null
mistralai/pixtral-12b-2409	Mistral Pixtral (2409)	Mistral Pixtral (2409)	Mistral Pixtral 12B is the first multimodal Mistral model for image understanding. ([blog](https://mistral.ai/news/pixtral-12b/))	Mistral AI	open	false	2024/9/17	null
mosaicml/mpt-30b	MPT (30B)	null	MPT (30B) is a Transformer trained from scratch on 1T tokens of text and code.	MosaicML	open	false	2023/6/22	30,000,000,000
mosaicml/mpt-30b-chat	MPT-Chat (30B)	null	MPT-Chat (30B) is a chatbot-like model for dialogue generation. It is built by finetuning MPT (30B), a Transformer trained from scratch on 1T tokens of text and code.	MosaicML	open	true	2023/6/22	30,000,000,000
mosaicml/mpt-7b	MPT (7B)	null	MPT (7B) is a Transformer trained from scratch on 1T tokens of text and code.	MosaicML	open	false	2023/5/5	6,700,000,000
mosaicml/mpt-7b-chat	MPT-Chat (7B)	null	MPT-Chat (7B) is a chatbot-like model for dialogue generation. It is built by finetuning MPT (30B) , a Transformer trained from scratch on 1T tokens of text and code.	MosaicML	open	true	2023/5/5	6,700,000,000
mosaicml/mpt-instruct-30b	MPT-Instruct (30B)	null	MPT-Instruct (30B) is a model for short-form instruction following. It is built by finetuning MPT (30B), a Transformer trained from scratch on 1T tokens of text and code.	MosaicML	open	false	2023/6/22	30,000,000,000
mosaicml/mpt-instruct-7b	MPT-Instruct (7B)	null	MPT-Instruct (7B) is a model for short-form instruction following. It is built by finetuning MPT (30B), a Transformer trained from scratch on 1T tokens of text and code.	MosaicML	open	false	2023/5/5	6,700,000,000
neurips/local	Local service	null	Local competition service	neurips	open	false	2021/12/1	1
NousResearch/Nous-Capybara-7B-V1.9	Nous Capybara 7B V1.9	Nous Capybara 7B V1.9	Nous Capybara 7B V1.9 is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-capybara-7b-v1-9/))	Nous Research	open	false	2024/4/18	7,000,000,000
NousResearch/Nous-Hermes-2-Mistral-7B-DPO	Nous Hermes 2 Mistral 7B DPO	Nous Hermes 2 Mistral 7B DPO	Nous Hermes 2 Mistral 7B DPO is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-2-mistral-7b-dpo/))	Nous Research	open	false	2024/4/18	7,000,000,000
NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO	Nous Hermes 2 Mixtral 8x7B DPO	Nous Hermes 2 Mixtral 8x7B DPO	Nous Hermes 2 Mixtral 8x7B DPO is a multimodal model trained on 8x7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-2-mixtral-8x7b-dpo/))	Nous Research	open	false	2024/4/18	46,700,000,000
NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT	Nous Hermes 2 Mixtral 8x7B SFT	Nous Hermes 2 Mixtral 8x7B SFT	Nous Hermes 2 Mixtral 8x7B SFT is a multimodal model trained on 8x7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-2-mixtral-8x7b-sft/))	Nous Research	open	false	2024/4/18	46,700,000,000
NousResearch/Nous-Hermes-2-Yi-34B	Nous Hermes 2 Yi 34B	Nous Hermes 2 Yi 34B	Nous Hermes 2 Yi 34B is a multimodal model trained on 34B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-2-yi-34b/))	Nous Research	open	false	2024/4/18	34,000,000,000
NousResearch/Nous-Hermes-Llama-2-7b	Nous Hermes Llama 2 7B	Nous Hermes Llama 2 7B	Nous Hermes Llama 2 7B is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-llama-2-7b/))	Nous Research	open	false	2024/4/18	7,000,000,000
NousResearch/Nous-Hermes-Llama2-13b	Nous Hermes Llama 2 13B	Nous Hermes Llama 2 13B	Nous Hermes Llama 2 13B is a multimodal model trained on 13B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-llama-2-13b/))	Nous Research	open	false	2024/4/18	13,000,000,000
nvidia/megatron-gpt2	Megatron GPT2	null	GPT-2 implemented in Megatron-LM ([paper](https://arxiv.org/abs/1909.08053)).	NVIDIA	open	true	null	null
Open-Orca/Mistral-7B-OpenOrca	Mistral 7B OpenOrca	Mistral 7B OpenOrca	Mistral 7B OpenOrca is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://openorca.com/mistral-7b-openorca/))	Open Orca	open	false	2024/4/18	7,000,000,000
openai/ada	ada (350M)	null	Original GPT-3 (350M parameters) autoregressive language model ([paper](https://arxiv.org/pdf/2005.14165.pdf), [docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2020/5/28	350,000,000
openai/babbage	babbage (1.3B)	null	Original GPT-3 (1.3B parameters) autoregressive language model ([paper](https://arxiv.org/pdf/2005.14165.pdf), [docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2020/5/28	1,300,000,000
openai/code-cushman-001	code-cushman-001 (12B)	null	Codex-style model that is a stronger, multilingual version of the Codex (12B) model in the [Codex paper](https://arxiv.org/pdf/2107.03374.pdf).	OpenAI	limited	false	null	null
openai/code-davinci-001	code-davinci-001	null	code-davinci-001 model	OpenAI	limited	true	null	null
openai/code-davinci-002	code-davinci-002	null	Codex-style model that is designed for pure code-completion tasks ([docs](https://beta.openai.com/docs/models/codex)).	OpenAI	limited	false	null	null
openai/curie	curie (6.7B)	null	Original GPT-3 (6.7B parameters) autoregressive language model ([paper](https://arxiv.org/pdf/2005.14165.pdf), [docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2020/5/28	6,700,000,000
openai/davinci	davinci (175B)	null	Original GPT-3 (175B parameters) autoregressive language model ([paper](https://arxiv.org/pdf/2005.14165.pdf), [docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2020/5/28	175,000,000,000
openai/gpt-3.5-turbo-0125	GPT-3.5 Turbo (0125)	GPT-3.5 Turbo (0125)	Sibling model of text-davinci-003 that is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2024-01-25.	OpenAI	limited	false	2023/6/13	null
openai/gpt-3.5-turbo-0301	GPT-3.5 Turbo (0301)	GPT-3.5 Turbo (0301)	Sibling model of text-davinci-003 that is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2023-03-01.	OpenAI	limited	false	2023/3/1	null
openai/gpt-3.5-turbo-0613	GPT-3.5 Turbo (0613)	GPT-3.5 Turbo (0613)	Sibling model of text-davinci-003 that is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2023-06-13.	OpenAI	limited	false	2023/6/13	null
openai/gpt-3.5-turbo-1106	GPT-3.5 Turbo (1106)	GPT-3.5 Turbo (1106)	Sibling model of text-davinci-003 that is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2023-11-06.	OpenAI	limited	false	2024/1/25	null
openai/gpt-3.5-turbo-16k-0613	gpt-3.5-turbo-16k-0613	null	Sibling model of text-davinci-003 is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2023-06-13 with a longer context length of 16,384 tokens.	OpenAI	limited	false	2023/6/13	null
openai/gpt-4-0314	gpt-4-0314	null	GPT-4 is a large multimodal model (currently only accepting text inputs and emitting text outputs) that is optimized for chat but works well for traditional completions tasks. Snapshot of gpt-4 from March 14th 2023.	OpenAI	limited	false	2023/3/14	null
openai/gpt-4-0613	GPT-4 (0613)	GPT-4 (0613)	GPT-4 is a large multimodal model (currently only accepting text inputs and emitting text outputs) that is optimized for chat but works well for traditional completions tasks. Snapshot of gpt-4 from 2023-06-13.	OpenAI	limited	false	2023/6/13	null
openai/gpt-4-1106-preview	gpt-4-1106-preview	null	GPT-4 Turbo (preview) is a large multimodal model that is optimized for chat but works well for traditional completions tasks. The model is cheaper and faster than the original GPT-4 model. Preview snapshot from November 6, 2023.	OpenAI	limited	false	2023/11/6	null
openai/gpt-4-1106-vision-preview	GPT-4V (1106 preview)	GPT-4V (1106 preview)	GPT-4V is a large multimodal model that accepts both text and images and is optimized for chat ([model card](https://openai.com/research/gpt-4v-system-card)).	OpenAI	limited	false	2023/11/6	null
openai/gpt-4-32k-0314	gpt-4-32k-0314	null	GPT-4 is a large multimodal model (currently only accepting text inputs and emitting text outputs) that is optimized for chat but works well for traditional completions tasks. Snapshot of gpt-4 with a longer context length of 32,768 tokens from March 14th 2023.	OpenAI	limited	false	2023/3/14	null
openai/gpt-4-32k-0613	gpt-4-32k-0613	null	GPT-4 is a large multimodal model (currently only accepting text inputs and emitting text outputs) that is optimized for chat but works well for traditional completions tasks. Snapshot of gpt-4 with a longer context length of 32,768 tokens from 2023-06-13.	OpenAI	limited	false	2023/6/13	null
openai/gpt-4-turbo-2024-04-09	GPT-4 Turbo (2024-04-09)	GPT-4 Turbo (2024-04-09)	GPT-4 Turbo (2024-04-09) is a large multimodal model that is optimized for chat but works well for traditional completions tasks. The model is cheaper and faster than the original GPT-4 model. Snapshot from 2024-04-09.	OpenAI	limited	false	2024/4/9	null
openai/gpt-4-vision-preview	GPT-4V (1106 preview)	GPT-4V (1106 preview)	GPT-4V is a large multimodal model that accepts both text and images and is optimized for chat ([model card](https://openai.com/research/gpt-4v-system-card)).	OpenAI	limited	false	2023/11/6	null
openai/gpt-4o-2024-05-13	GPT-4o (2024-05-13)	GPT-4o (2024-05-13)	GPT-4o (2024-05-13) is a large multimodal model that accepts as input any combination of text, audio, and image and generates any combination of text, audio, and image outputs. ([blog](https://openai.com/index/hello-gpt-4o/))	OpenAI	limited	false	2024/4/9	null
openai/gpt-4o-2024-08-06	GPT-4o (2024-08-06)	GPT-4o (2024-08-06)	GPT-4o (2024-08-06) is a large multimodal model that accepts as input any combination of text, audio, and image and generates any combination of text, audio, and image outputs. ([blog](https://openai.com/index/introducing-structured-outputs-in-the-api/))	OpenAI	limited	false	2024/8/6	null
openai/gpt-4o-mini-2024-07-18	GPT-4o mini (2024-07-18)	GPT-4o mini (2024-07-18)	GPT-4o mini (2024-07-18) is a multimodal model with a context window of 128K tokens and improved handling of non-English text. ([blog](https://openai.com/index/gpt-4o-mini-advancing-cost-efficient-intelligence/))	OpenAI	limited	false	2024/7/18	null
openai/text-ada-001	text-ada-001	null	text-ada-001 model that involves supervised fine-tuning on human-written demonstrations ([docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2022/1/27	350,000,000
openai/text-babbage-001	text-babbage-001	null	text-babbage-001 model that involves supervised fine-tuning on human-written demonstrations ([docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2022/1/27	1,300,000,000
openai/text-curie-001	text-curie-001	null	text-curie-001 model that involves supervised fine-tuning on human-written demonstrations ([docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2022/1/27	6,700,000,000
openai/text-davinci-001	text-davinci-001	null	text-davinci-001 model that involves supervised fine-tuning on human-written demonstrations ([docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	true	2022/1/27	175,000,000,000
openai/text-davinci-002	text-davinci-002	null	text-davinci-002 model that involves supervised fine-tuning on human-written demonstrations. Derived from code-davinci-002 ([docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2022/1/27	175,000,000,000
openai/text-davinci-003	text-davinci-003	null	text-davinci-003 model that involves reinforcement learning (PPO) with reward models. Derived from text-davinci-002 ([docs](https://beta.openai.com/docs/model-index-for-researchers)).	OpenAI	limited	false	2022/11/28	175,000,000,000
openchat/openchat-3.5-1210	OpenChat 3.5 (1210)	OpenChat 3.5 (1210)	OpenChat 3.5 is a large language model trained on 1210 billion parameters. ([blog](https://openchat.com/openchat-3.5-1210/))	OpenChat	open	false	2024/4/18	1,210,000,000,000
openthaigpt/openthaigpt-1.0.0-13b-chat	OpenThaiGPT v1.0.0 (13B)	OpenThaiGPT v1.0.0 (13B)	OpenThaiGPT v1.0.0 (13B) is a Thai language chat model based on Llama 2 that has been specifically fine-tuned for Thai instructions and enhanced by incorporating over 10,000 of the most commonly used Thai words into the dictionary. ([blog post](https://openthaigpt.aieat.or.th/openthaigpt-1.0.0-less-than-8-apr-2024-greater-than))	OpenThaiGPT	open	false	2024/4/8	13,000,000,000

huggingface/santacoder

SantaCoder (1.1B)

null

SantaCoder (1.1B parameters) model trained on the Python, Java, and JavaScript subset of The Stack (v1.1) ([model card](https://huggingface.co/bigcode/santacoder)).

BigCode

open

false

null

huggingface/starcoder

StarCoder (15.5B)

null

The StarCoder (15.5B parameter) model trained on 80+ programming languages from The Stack (v1.2) ([model card](https://huggingface.co/bigcode/starcoder)).

BigCode

open

false

null

HuggingFaceM4/idefics-80b

IDEFICS (80B)

null

IDEFICS (80B parameters) is an open-source model based on DeepMind's Flamingo. ([blog](https://huggingface.co/blog/idefics))

HuggingFace

open

false

2023/8/22

80,000,000,000

HuggingFaceM4/idefics-80b-instruct

IDEFICS instruct (80B)

null

IDEFICS instruct (80B parameters) is an open-source model based on DeepMind's Flamingo. ([blog](https://huggingface.co/blog/idefics))

HuggingFace

open

false

2023/8/22

80,000,000,000

HuggingFaceM4/idefics-9b

IDEFICS (9B)

null

IDEFICS (9B parameters) is an open-source model based on DeepMind's Flamingo. ([blog](https://huggingface.co/blog/idefics))

HuggingFace

open

false

2023/8/22

9,000,000,000

HuggingFaceM4/idefics-9b-instruct

IDEFICS instruct (9B)

null

IDEFICS instruct (9B parameters) is an open-source model based on DeepMind's Flamingo. ([blog](https://huggingface.co/blog/idefics))

HuggingFace

open

false

2023/8/22

9,000,000,000

HuggingFaceM4/idefics2-8b

IDEFICS 2 (8B)

IDEFICS 2 (8B parameters) is an open multimodal model that accepts arbitrary sequences of image and text inputs and produces text outputs. ([blog](https://huggingface.co/blog/idefics2)).

HuggingFace

open

false

2024/4/15

8,000,000,000

lightningai/lit-gpt

Lit-GPT

null

Lit-GPT is an optimized collection of open-source LLMs for finetuning and inference. It supports – Falcon, Llama 2, Vicuna, LongChat, and other top-performing open-source large language models.

Lightning AI

open

false

2023/4/4

1

lmsys/vicuna-13b-v1.3

Vicuna v1.3 (13B)

null

Vicuna v1.3 (13B) is an open-source chatbot trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.

LMSYS

open

false

2023/6/22

13,000,000,000

lmsys/vicuna-13b-v1.5

Vicuna 13B (v1.5)

Vicuna is a large language model trained on 13 billion parameters. ([blog](https://lmsys.com/vicuna-13b-v1.5/))

LMSYS

open

false

2024/4/18

13,000,000,000

lmsys/vicuna-7b-v1.3

Vicuna v1.3 (7B)

null

Vicuna v1.3 (7B) is an open-source chatbot trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.

LMSYS

open

false

2023/6/22

7,000,000,000

lmsys/vicuna-7b-v1.5

Vicuna 7B (v1.5)

Vicuna is a large language model trained on 7 billion parameters. ([blog](https://lmsys.com/vicuna-7b-v1.5/))

LMSYS

open

false

2024/4/18

7,000,000,000

meta-llama/Llama-2-13b-chat-hf

Llama 2 13B Chat

Llama 2 13B Chat is a large language model trained on 13 billion parameters. ([blog](https://meta-llama.com/llama-2-13b-chat-hf/))

Meta Llama

open

false

2024/4/18

13,000,000,000

meta-llama/Llama-2-70b-chat-hf

Llama 2 70B Chat

Llama 2 70B Chat is a large language model trained on 70 billion parameters. ([blog](https://meta-llama.com/llama-2-70b-chat-hf/))

Meta Llama

open

false

2024/4/18

70,000,000,000

meta-llama/Llama-2-7b-chat-hf

Llama 2 7B Chat

Llama 2 7B Chat is a large language model trained on 7 billion parameters. ([blog](https://meta-llama.com/llama-2-7b-chat-hf/))

Meta Llama

open

false

2024/4/18

7,000,000,000

meta/llama-13b

LLaMA (13B)

null

LLaMA is a collection of foundation language models ranging from 7B to 65B parameters.

Meta

open

false

2023/2/24

13,000,000,000

meta/llama-2-13b

Llama 2 (13B)

null

Llama 2 pretrained models are trained on 2 trillion tokens, and have double the context length than Llama 1.

Meta

open

false

2023/7/18

13,000,000,000

meta/llama-2-70b

Llama 2 (70B)

null

Llama 2 pretrained models are trained on 2 trillion tokens, and have double the context length than Llama 1.

Meta

open

false

2023/7/18

70,000,000,000

meta/llama-2-7b

Llama 2 (7B)

null

Llama 2 pretrained models are trained on 2 trillion tokens, and have double the context length than Llama 1.

Meta

open

false

2023/7/18

7,000,000,000

meta/llama-3-70b

Llama 3 (70B)

Llama 3 is a family of language models that have been trained on more than 15 trillion tokens, and use Grouped-Query Attention (GQA) for improved inference scalability. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)

Meta

open

false

2024/4/18

70,000,000,000

meta/llama-3-70b-chat

Llama 3 Instruct (70B)

Llama 3 is a family of language models that have been trained on more than 15 trillion tokens, and use Grouped-Query Attention (GQA) for improved inference scalability. It used SFT, rejection sampling, PPO and DPO for post-training. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)

Meta

open

false

2024/4/18

70,000,000,000

meta/llama-3-8b

Llama 3 (8B)

Llama 3 is a family of language models that have been trained on more than 15 trillion tokens, and use Grouped-Query Attention (GQA) for improved inference scalability. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)

Meta

open

false

2024/4/18

8,000,000,000

meta/llama-3-8b-chat

Llama 3 Instruct (8B)

Llama 3 is a family of language models that have been trained on more than 15 trillion tokens, and use Grouped-Query Attention (GQA) for improved inference scalability. It used SFT, rejection sampling, PPO and DPO for post-training. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)

Meta

open

false

2024/4/18

8,000,000,000

meta/llama-3.1-405b-instruct-turbo

Llama 3.1 Instruct Turbo (405B)

Llama 3.1 (405B) is part of the Llama 3 family of dense Transformer models that that natively support multilinguality, coding, reasoning, and tool usage. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/), [blog](https://ai.meta.com/blog/meta-llama-3-1/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))

Meta

open

false

2024/7/23

405,000,000,000

meta/llama-3.1-70b-instruct-turbo

Llama 3.1 Instruct Turbo (70B)

Llama 3.1 (70B) is part of the Llama 3 family of dense Transformer models that that natively support multilinguality, coding, reasoning, and tool usage. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/), [blog](https://ai.meta.com/blog/meta-llama-3-1/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))

Meta

open

false

2024/7/23

70,000,000,000

meta/llama-3.1-8b-instruct-turbo

Llama 3.1 Instruct Turbo (8B)

Llama 3.1 (8B) is part of the Llama 3 family of dense Transformer models that that natively support multilinguality, coding, reasoning, and tool usage. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/), [blog](https://ai.meta.com/blog/meta-llama-3-1/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))

Meta

open

false

2024/7/23

8,000,000,000

meta/llama-3.2-11b-vision-instruct-turbo

Llama 3.2 Vision Instruct Turbo (11B)

The Llama 3.2 Vision collection of multimodal large language models (LLMs) is a collection of pretrained and instruction-tuned image reasoning generative models in 11B and 90B sizes. ([blog](https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))

Meta

open

false

2024/9/25

10,700,000,000

meta/llama-3.2-90b-vision-instruct-turbo

Llama 3.2 Vision Instruct Turbo (90B)

The Llama 3.2 Vision collection of multimodal large language models (LLMs) is a collection of pretrained and instruction-tuned image reasoning generative models in 11B and 90B sizes. ([blog](https://ai.meta.com/blog/llama-3-2-connect-2024-vision-edge-mobile-devices/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))

Meta

open

false

2024/9/25

88,600,000,000

meta/llama-3.3-70b-instruct-turbo

Llama 3.3 Instruct Turbo (70B)

Llama 3.3 (70B) is part of the Llama 3 family of dense Transformer models that that natively support multilinguality, coding, reasoning, and tool usage. ([paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/)) Turbo is Together's implementation, providing a near negligible difference in quality from the reference implementation with faster performance and lower cost, currently using FP8 quantization. ([blog](https://www.together.ai/blog/llama-31-quality))

Meta

open

false

2024/12/6

70,000,000,000

meta/llama-30b

LLaMA (30B)

null

LLaMA is a collection of foundation language models ranging from 7B to 65B parameters.

Meta

open

false

2023/2/24

30,000,000,000

meta/llama-65b

LLaMA (65B)

null

LLaMA is a collection of foundation language models ranging from 7B to 65B parameters.

Meta

open

false

2023/2/24

65,000,000,000

meta/llama-7b

LLaMA (7B)

null

LLaMA is a collection of foundation language models ranging from 7B to 65B parameters.

Meta

open

false

2023/2/24

7,000,000,000

microsoft/llava-1.5-13b-hf

LLaVA 1.5 (13B)

LLaVa is an open-source chatbot trained by fine-tuning LlamA/Vicuna on GPT-generated multimodal instruction-following data. ([paper](https://arxiv.org/abs/2304.08485))

Microsoft

open

false

2023/10/5

13,000,000,000

microsoft/phi-2

Phi-2

Phi-2 is a Transformer with 2.7 billion parameters. It was trained using the same data sources as Phi-1.5, augmented with a new data source that consists of various NLP synthetic texts and filtered websites (for safety and educational value)

Microsoft

open

false

2023/10/5

13,000,000,000

microsoft/phi-3-medium-4k-instruct

Phi-3 (14B)

Phi-3-Medium-4K-Instruct is a lightweight model trained with synthetic data and filtered publicly available website data with a focus on high-quality and reasoning dense properties. ([paper](https://arxiv.org/abs/2404.14219), [blog](https://azure.microsoft.com/en-us/blog/new-models-added-to-the-phi-3-family-available-on-microsoft-azure/))

Microsoft

open

false

2024/5/21

14,000,000,000

microsoft/phi-3-small-8k-instruct

Phi-3 (7B)

Phi-3-Small-8K-Instruct is a lightweight model trained with synthetic data and filtered publicly available website data with a focus on high-quality and reasoning dense properties. ([paper](https://arxiv.org/abs/2404.14219), [blog](https://azure.microsoft.com/en-us/blog/new-models-added-to-the-phi-3-family-available-on-microsoft-azure/))

Microsoft

open

false

2024/5/21

7,000,000,000

microsoft/TNLGv2_530B

TNLG v2 (530B)

null

TNLG v2 (530B parameters) autoregressive language model trained on a filtered subset of the Pile and CommonCrawl ([paper](https://arxiv.org/pdf/2201.11990.pdf)).

Microsoft/NVIDIA

closed

false

2022/1/28

530,000,000,000

microsoft/TNLGv2_7B

TNLG v2 (6.7B)

null

TNLG v2 (6.7B parameters) autoregressive language model trained on a filtered subset of the Pile and CommonCrawl ([paper](https://arxiv.org/pdf/2201.11990.pdf)).

Microsoft/NVIDIA

closed

false

2022/1/28

6,700,000,000

mistralai/mistral-7b-instruct-v0.1

Mistral Instruct v0.1 (7B)

Mistral v0.1 Instruct 7B is a 7.3B parameter transformer model that uses Grouped-Query Attention (GQA) and Sliding-Window Attention (SWA). The instruct version was fined-tuned using publicly available conversation datasets. ([blog post](https://mistral.ai/news/announcing-mistral-7b/))

Mistral AI

open

false

2023/9/27

7,300,000,000

mistralai/Mistral-7B-Instruct-v0.2

Mistral 7B Instruct v0.2

Mistral 7B Instruct v0.2 is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://mistral.ai/mistral-7b-instruct-v0-2/))

Mistral AI

open

false

2024/4/18

7,000,000,000

mistralai/mistral-7b-instruct-v0.3

Mistral Instruct v0.3 (7B)

Mistral v0.3 Instruct 7B is a 7.3B parameter transformer model that uses Grouped-Query Attention (GQA). Compared to v0.1, v0.2 has a 32k context window and no Sliding-Window Attention (SWA). ([blog post](https://mistral.ai/news/la-plateforme/))

Mistral AI

open

false

2024/5/22

7,300,000,000

mistralai/mistral-7b-v0.1

Mistral v0.1 (7B)

null

Mistral 7B is a 7.3B parameter transformer model that uses Grouped-Query Attention (GQA) and Sliding-Window Attention (SWA).

Mistral AI

open

false

2023/9/27

7,300,000,000

mistralai/Mistral-7B-v0.1

Mistral 7B v0.1

Mistral 7B v0.1 is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://mistral.ai/mistral-7b-v0-1/))

Mistral AI

open

false

2024/4/18

7,000,000,000

mistralai/mistral-large-2402

Mistral Large (2402)

Mistral Large is a multilingual model with a 32K tokens context window and function-calling capabilities. ([blog](https://mistral.ai/news/mistral-large/))

Mistral AI

limited

false

2023/2/26

null

mistralai/mistral-large-2407

Mistral Large 2 (2407)

Mistral Large 2 is a 123 billion parameter model that has a 128k context window and supports dozens of languages and 80+ coding languages. ([blog](https://mistral.ai/news/mistral-large-2407/))

Mistral AI

open

false

2023/7/24

123,000,000,000

mistralai/mistral-medium-2312

Mistral Medium (2312)

Mistral is a transformer model that uses Grouped-Query Attention (GQA) and Sliding-Window Attention (SWA).

Mistral AI

limited

false

2023/12/11

null

mistralai/mistral-small-2402

Mistral Small (2402)

Mistral Small is a multilingual model with a 32K tokens context window and function-calling capabilities. ([blog](https://mistral.ai/news/mistral-large/))

Mistral AI

limited

false

2023/2/26

null

mistralai/mixtral-8x22b

Mixtral (8x22B)

Mistral AI's mixture-of-experts model that uses 39B active parameters out of 141B ([blog post](https://mistral.ai/news/mixtral-8x22b/)).

Mistral AI

open

false

2024/4/10

176,000,000,000

mistralai/mixtral-8x22b-instruct-v0.1

Mixtral Instruct (8x22B)

Mistral AI's mixture-of-experts model that uses 39B active parameters out of 141B ([blog post](https://mistral.ai/news/mixtral-8x22b/)).

Mistral AI

open

false

2024/4/10

176,000,000,000

mistralai/mixtral-8x7b-32kseqlen

Mixtral (8x7B 32K seqlen)

Mixtral is a mixture-of-experts model that has 46.7B total parameters but only uses 12.9B parameters per token. ([blog post](https://mistral.ai/news/mixtral-of-experts/), [tweet](https://twitter.com/MistralAI/status/1733150512395038967)).

Mistral AI

open

false

2023/12/8

46,700,000,000

mistralai/mixtral-8x7b-instruct-v0.1

Mixtral Instruct (8x7B)

Mixtral Instruct (8x7B) is a version of Mixtral (8x7B) that was optimized through supervised fine-tuning and direct preference optimisation (DPO) for careful instruction following. ([blog post](https://mistral.ai/news/mixtral-of-experts/)).

Mistral AI

open

false

2023/12/11

46,700,000,000

mistralai/open-mistral-nemo-2407

Mistral NeMo (2402)

Mistral NeMo is a multilingual 12B model with a large context window of 128K tokens. ([blog](https://mistral.ai/news/mistral-nemo/))

Mistral AI

open

false

2024/7/18

null

mistralai/pixtral-12b-2409

Mistral Pixtral (2409)

Mistral Pixtral 12B is the first multimodal Mistral model for image understanding. ([blog](https://mistral.ai/news/pixtral-12b/))

Mistral AI

open

false

2024/9/17

null

mosaicml/mpt-30b

MPT (30B)

null

MPT (30B) is a Transformer trained from scratch on 1T tokens of text and code.

MosaicML

open

false

2023/6/22

30,000,000,000

mosaicml/mpt-30b-chat

MPT-Chat (30B)

null

MPT-Chat (30B) is a chatbot-like model for dialogue generation. It is built by finetuning MPT (30B), a Transformer trained from scratch on 1T tokens of text and code.

MosaicML

open

true

2023/6/22

30,000,000,000

mosaicml/mpt-7b

MPT (7B)

null

MPT (7B) is a Transformer trained from scratch on 1T tokens of text and code.

MosaicML

open

false

2023/5/5

6,700,000,000

mosaicml/mpt-7b-chat

MPT-Chat (7B)

null

MPT-Chat (7B) is a chatbot-like model for dialogue generation. It is built by finetuning MPT (30B) , a Transformer trained from scratch on 1T tokens of text and code.

MosaicML

open

true

2023/5/5

6,700,000,000

mosaicml/mpt-instruct-30b

MPT-Instruct (30B)

null

MPT-Instruct (30B) is a model for short-form instruction following. It is built by finetuning MPT (30B), a Transformer trained from scratch on 1T tokens of text and code.

MosaicML

open

false

2023/6/22

30,000,000,000

mosaicml/mpt-instruct-7b

MPT-Instruct (7B)

null

MPT-Instruct (7B) is a model for short-form instruction following. It is built by finetuning MPT (30B), a Transformer trained from scratch on 1T tokens of text and code.

MosaicML

open

false

2023/5/5

6,700,000,000

neurips/local

Local service

null

Local competition service

neurips

open

false

2021/12/1

1

NousResearch/Nous-Capybara-7B-V1.9

Nous Capybara 7B V1.9

Nous Capybara 7B V1.9 is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-capybara-7b-v1-9/))

Nous Research

open

false

2024/4/18

7,000,000,000

NousResearch/Nous-Hermes-2-Mistral-7B-DPO

Nous Hermes 2 Mistral 7B DPO

Nous Hermes 2 Mistral 7B DPO is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-2-mistral-7b-dpo/))

Nous Research

open

false

2024/4/18

7,000,000,000

NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO

Nous Hermes 2 Mixtral 8x7B DPO

Nous Hermes 2 Mixtral 8x7B DPO is a multimodal model trained on 8x7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-2-mixtral-8x7b-dpo/))

Nous Research

open

false

2024/4/18

46,700,000,000

NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT

Nous Hermes 2 Mixtral 8x7B SFT

Nous Hermes 2 Mixtral 8x7B SFT is a multimodal model trained on 8x7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-2-mixtral-8x7b-sft/))

Nous Research

open

false

2024/4/18

46,700,000,000

NousResearch/Nous-Hermes-2-Yi-34B

Nous Hermes 2 Yi 34B

Nous Hermes 2 Yi 34B is a multimodal model trained on 34B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-2-yi-34b/))

Nous Research

open

false

2024/4/18

34,000,000,000

NousResearch/Nous-Hermes-Llama-2-7b

Nous Hermes Llama 2 7B

Nous Hermes Llama 2 7B is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-llama-2-7b/))

Nous Research

open

false

2024/4/18

7,000,000,000

NousResearch/Nous-Hermes-Llama2-13b

Nous Hermes Llama 2 13B

Nous Hermes Llama 2 13B is a multimodal model trained on 13B parameters with a 32K token sequence length. ([blog](https://nousresearch.com/nous-hermes-llama-2-13b/))

Nous Research

open

false

2024/4/18

13,000,000,000

nvidia/megatron-gpt2

Megatron GPT2

null

GPT-2 implemented in Megatron-LM ([paper](https://arxiv.org/abs/1909.08053)).

NVIDIA

open

true

null

Open-Orca/Mistral-7B-OpenOrca

Mistral 7B OpenOrca

Mistral 7B OpenOrca is a multimodal model trained on 7B parameters with a 32K token sequence length. ([blog](https://openorca.com/mistral-7b-openorca/))

Open Orca

open

false

2024/4/18

7,000,000,000

openai/ada

ada (350M)

null

Original GPT-3 (350M parameters) autoregressive language model ([paper](https://arxiv.org/pdf/2005.14165.pdf), [docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2020/5/28

350,000,000

openai/babbage

babbage (1.3B)

null

Original GPT-3 (1.3B parameters) autoregressive language model ([paper](https://arxiv.org/pdf/2005.14165.pdf), [docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2020/5/28

1,300,000,000

openai/code-cushman-001

code-cushman-001 (12B)

null

Codex-style model that is a stronger, multilingual version of the Codex (12B) model in the [Codex paper](https://arxiv.org/pdf/2107.03374.pdf).

OpenAI

limited

false

null

openai/code-davinci-001

code-davinci-001

null

code-davinci-001 model

OpenAI

limited

true

null

openai/code-davinci-002

code-davinci-002

null

Codex-style model that is designed for pure code-completion tasks ([docs](https://beta.openai.com/docs/models/codex)).

OpenAI

limited

false

null

openai/curie

curie (6.7B)

null

Original GPT-3 (6.7B parameters) autoregressive language model ([paper](https://arxiv.org/pdf/2005.14165.pdf), [docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2020/5/28

6,700,000,000

openai/davinci

davinci (175B)

null

Original GPT-3 (175B parameters) autoregressive language model ([paper](https://arxiv.org/pdf/2005.14165.pdf), [docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2020/5/28

175,000,000,000

openai/gpt-3.5-turbo-0125

GPT-3.5 Turbo (0125)

Sibling model of text-davinci-003 that is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2024-01-25.

OpenAI

limited

false

2023/6/13

null

openai/gpt-3.5-turbo-0301

GPT-3.5 Turbo (0301)

Sibling model of text-davinci-003 that is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2023-03-01.

OpenAI

limited

false

2023/3/1

null

openai/gpt-3.5-turbo-0613

GPT-3.5 Turbo (0613)

Sibling model of text-davinci-003 that is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2023-06-13.

OpenAI

limited

false

2023/6/13

null

openai/gpt-3.5-turbo-1106

GPT-3.5 Turbo (1106)

Sibling model of text-davinci-003 that is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2023-11-06.

OpenAI

limited

false

2024/1/25

null

openai/gpt-3.5-turbo-16k-0613

gpt-3.5-turbo-16k-0613

null

Sibling model of text-davinci-003 is optimized for chat but works well for traditional completions tasks as well. Snapshot from 2023-06-13 with a longer context length of 16,384 tokens.

OpenAI

limited

false

2023/6/13

null

openai/gpt-4-0314

gpt-4-0314

null

GPT-4 is a large multimodal model (currently only accepting text inputs and emitting text outputs) that is optimized for chat but works well for traditional completions tasks. Snapshot of gpt-4 from March 14th 2023.

OpenAI

limited

false

2023/3/14

null

openai/gpt-4-0613

GPT-4 (0613)

GPT-4 is a large multimodal model (currently only accepting text inputs and emitting text outputs) that is optimized for chat but works well for traditional completions tasks. Snapshot of gpt-4 from 2023-06-13.

OpenAI

limited

false

2023/6/13

null

openai/gpt-4-1106-preview

gpt-4-1106-preview

null

GPT-4 Turbo (preview) is a large multimodal model that is optimized for chat but works well for traditional completions tasks. The model is cheaper and faster than the original GPT-4 model. Preview snapshot from November 6, 2023.

OpenAI

limited

false

2023/11/6

null

openai/gpt-4-1106-vision-preview

GPT-4V (1106 preview)

GPT-4V is a large multimodal model that accepts both text and images and is optimized for chat ([model card](https://openai.com/research/gpt-4v-system-card)).

OpenAI

limited

false

2023/11/6

null

openai/gpt-4-32k-0314

gpt-4-32k-0314

null

GPT-4 is a large multimodal model (currently only accepting text inputs and emitting text outputs) that is optimized for chat but works well for traditional completions tasks. Snapshot of gpt-4 with a longer context length of 32,768 tokens from March 14th 2023.

OpenAI

limited

false

2023/3/14

null

openai/gpt-4-32k-0613

gpt-4-32k-0613

null

GPT-4 is a large multimodal model (currently only accepting text inputs and emitting text outputs) that is optimized for chat but works well for traditional completions tasks. Snapshot of gpt-4 with a longer context length of 32,768 tokens from 2023-06-13.

OpenAI

limited

false

2023/6/13

null

openai/gpt-4-turbo-2024-04-09

GPT-4 Turbo (2024-04-09)

GPT-4 Turbo (2024-04-09) is a large multimodal model that is optimized for chat but works well for traditional completions tasks. The model is cheaper and faster than the original GPT-4 model. Snapshot from 2024-04-09.

OpenAI

limited

false

2024/4/9

null

openai/gpt-4-vision-preview

GPT-4V (1106 preview)

GPT-4V is a large multimodal model that accepts both text and images and is optimized for chat ([model card](https://openai.com/research/gpt-4v-system-card)).

OpenAI

limited

false

2023/11/6

null

openai/gpt-4o-2024-05-13

GPT-4o (2024-05-13)

GPT-4o (2024-05-13) is a large multimodal model that accepts as input any combination of text, audio, and image and generates any combination of text, audio, and image outputs. ([blog](https://openai.com/index/hello-gpt-4o/))

OpenAI

limited

false

2024/4/9

null

openai/gpt-4o-2024-08-06

GPT-4o (2024-08-06)

GPT-4o (2024-08-06) is a large multimodal model that accepts as input any combination of text, audio, and image and generates any combination of text, audio, and image outputs. ([blog](https://openai.com/index/introducing-structured-outputs-in-the-api/))

OpenAI

limited

false

2024/8/6

null

openai/gpt-4o-mini-2024-07-18

GPT-4o mini (2024-07-18)

GPT-4o mini (2024-07-18) is a multimodal model with a context window of 128K tokens and improved handling of non-English text. ([blog](https://openai.com/index/gpt-4o-mini-advancing-cost-efficient-intelligence/))

OpenAI

limited

false

2024/7/18

null

openai/text-ada-001

text-ada-001

null

text-ada-001 model that involves supervised fine-tuning on human-written demonstrations ([docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2022/1/27

350,000,000

openai/text-babbage-001

text-babbage-001

null

text-babbage-001 model that involves supervised fine-tuning on human-written demonstrations ([docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2022/1/27

1,300,000,000

openai/text-curie-001

text-curie-001

null

text-curie-001 model that involves supervised fine-tuning on human-written demonstrations ([docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2022/1/27

6,700,000,000

openai/text-davinci-001

text-davinci-001

null

text-davinci-001 model that involves supervised fine-tuning on human-written demonstrations ([docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

true

2022/1/27

175,000,000,000

openai/text-davinci-002

text-davinci-002

null

text-davinci-002 model that involves supervised fine-tuning on human-written demonstrations. Derived from code-davinci-002 ([docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2022/1/27

175,000,000,000

openai/text-davinci-003

text-davinci-003

null

text-davinci-003 model that involves reinforcement learning (PPO) with reward models. Derived from text-davinci-002 ([docs](https://beta.openai.com/docs/model-index-for-researchers)).

OpenAI

limited

false

2022/11/28

175,000,000,000

openchat/openchat-3.5-1210

OpenChat 3.5 (1210)

OpenChat 3.5 is a large language model trained on 1210 billion parameters. ([blog](https://openchat.com/openchat-3.5-1210/))

OpenChat

open

false

2024/4/18

1,210,000,000,000

openthaigpt/openthaigpt-1.0.0-13b-chat

OpenThaiGPT v1.0.0 (13B)

OpenThaiGPT v1.0.0 (13B) is a Thai language chat model based on Llama 2 that has been specifically fine-tuned for Thai instructions and enhanced by incorporating over 10,000 of the most commonly used Thai words into the dictionary. ([blog post](https://openthaigpt.aieat.or.th/openthaigpt-1.0.0-less-than-8-apr-2024-greater-than))

OpenThaiGPT

open

false

2024/4/8

13,000,000,000