Submit Tool
#

Legacy

AI tools tagged Legacy.

100+ tools · handpicked & curated
// price
abab6.5 MiniMax's proprietary flagship model prior to its open-weight pivot
ALBERT Google's parameter-sharing variant of BERT designed for efficiency at scale
AMD Radeon Instinct MI25 AMD's first Instinct-branded datacenter accelerator
AMD Radeon Instinct MI60 Vega-based datacenter accelerator, AMD's first 7nm GPU
AMD Radeon RX Vega 64 Flagship Vega-generation consumer GPU with HBM2
Aquila BAAI's first open bilingual (Chinese/English) foundation model
AudioLM Google's framework for generating realistic speech and audio continuations
Baichuan 2 Baichuan's open-weight bilingual foundation model
Baichuan-13B Baichuan's open-weight bilingual model, predating the Baichuan 2/3/4 generations
BERT Google's 2018 bidirectional encoder that reshaped NLP research
BioGPT Microsoft's domain-specific generative model pre-trained on biomedical literature
BlenderBot 3 Meta's open-domain chatbot research model that browses the web to stay current
BLIP Salesforce's vision-language pretraining model for captioning and retrieval
BLOOM The open, multilingual model built by the BigScience collaboration
BLOOMZ An instruction-tuned variant of BLOOM trained to follow cross-lingual instructions
ByT5 Google's token-free variant of T5 that operates directly on raw bytes
ChatGLM3 Zhipu's third-generation open-weight bilingual dialogue model
Chinchilla DeepMind's 2022 research model that reset scaling-law assumptions for LLM training
Claude 1 Anthropic's first publicly released Claude model
Claude 2 Anthropic's second-generation Claude, ahead of the Claude 2.1 refresh
Claude 2.1 Anthropic's 2023 model that pushed context length to 200K tokens
Claude 3 Haiku Anthropic's fastest model in the original Claude 3 family
Claude 3 Opus Anthropic's 2024 flagship, the top tier of the original Claude 3 family
Claude 3 Sonnet The balanced mid-tier of Anthropic's original Claude 3 lineup
Claude Instant Anthropic's original fast, low-cost model tier
CLIP OpenAI's foundational image-text contrastive model that underlies much of modern multimodal AI
Codex OpenAI's original code-completion model that powered early GitHub Copilot
Command Light Cohere's smallest, fastest Command tier
DALL-E OpenAI's original 2021 text-to-image model
DALL-E 2 OpenAI's second image generator, the first to bring photorealistic text-to-image to the mainstream
DALL-E 3 OpenAI's image generation model integrated directly into ChatGPT
DeepFloyd IF A pixel-space (non-latent) text-to-image model developed by the Stability-backed DeepFloyd lab
DeepSeek LLM 67B DeepSeek's first dense open-weight foundation model
DeepSeek-Coder-33B DeepSeek's earlier flagship code model, preceding DeepSeek-Coder-V2
DeepSeek-MoE-16B DeepSeek's early fine-grained mixture-of-experts model that informed its later MoE architectures
DeepSeek-V2 DeepSeek's 2024 mixture-of-experts model that undercut rivals on API price
DeepSeek-VL DeepSeek's first vision-language model, preceding DeepSeek-VL2
DistilBERT Hugging Face's distilled, 40%-smaller version of BERT that retains most of its performance
Dolly 2.0 Databricks' first fully open, commercially-usable instruction-following model
ELECTRA Google's more sample-efficient pretraining approach using a replaced-token-detection objective
ELMo AI2's deep contextualized word representation model that predated the transformer era
Emu Meta's photorealistic text-to-image model behind Meta AI's image features
Emu Video Meta's text-to-video model built on the Emu image generator
ERNIE 4.0 Baidu's flagship model prior to the 4.5 generation
EXAONE 3.0 LG's earlier EXAONE generation, preceding EXAONE 3.5
Falcon 180B TII's open-weight model that led open leaderboards on release
Falcon 40B TII's earlier open-weight model that led leaderboards before Falcon 180B
Falcon 7B TII's compact Falcon tier that helped popularize the original Falcon release
Flan-T5 Google's instruction-tuned, open-weight successor to T5
Galactica Meta's 2022 model trained on scientific literature, pulled days after launch
Gemini 1.0 Pro The mid-tier of Google's first Gemini generation
Gemini 1.0 Ultra Google's first Gemini-generation flagship, launched at the start of 2024
Gemini 1.5 Pro Google's 2024 model that introduced the 1M-token context window
Gemma Google's original open-weight Gemma release, built from the same research as Gemini
Google TPU v1 Google's first-generation inference-only TPU, never sold externally
Gopher DeepMind's 280B-parameter research model that informed the Chinchilla scaling laws
GPT-1 OpenAI's original 2018 paper model that introduced the GPT architecture
GPT-2 OpenAI's 2019 model, once withheld from release over misuse concerns
GPT-3 The 2020 model that first showed large-scale few-shot learning was possible
GPT-3 Davinci The largest of the original GPT-3 model sizes
GPT-3.5 Turbo The model that powered ChatGPT's original public launch
GPT-4 The 2023 release that reset expectations for what LLMs could do
GPT-J EleutherAI's early open-weight GPT-3-style model
GPT-Neo 2.7B EleutherAI's early GPT-3-style open model, a predecessor to GPT-J and GPT-NeoX
GPT-NeoX-20B EleutherAI's open-weight model, one of the largest public checkpoints of its era
Granite 13B IBM's earlier enterprise foundation model, preceding the Granite 3 generation
Grok-1 xAI's first model, later open-weighted under Apache 2.0
Grok-1.5 xAI's transitional model that added long-context reasoning ahead of Grok 2
Grok-1.5V xAI's first vision-capable Grok model, announced as a preview ahead of wider release
Ideogram 1.0 Ideogram's first model, notable for unusually reliable in-image text rendering
Imagen Google's original text-to-image diffusion model, introduced alongside Parti as a research preview
Imagen 2 Google's second-generation text-to-image model with improved photorealism and text rendering
Inflection-1 Inflection's original model that first powered the Pi assistant
InstructGPT OpenAI's 2022 model that introduced RLHF instruction-tuning at scale
Jina Embeddings v2 Jina AI's embedding model, notable for an 8k-token context window at launch
Jurassic-1 AI21's first large language model, a GPT-3 era contemporary
Jurassic-2 AI21's earlier proprietary large language model line
Jurassic-2 Mid The mid-sized tier of AI21's Jurassic-2 family, preceding the Jamba architecture switch
LaMDA Google's 2021 conversational model that underpinned the original Bard
Leonardo Diffusion XL Leonardo's earlier fine-tuned Stable Diffusion XL base model
LLaMA Meta's original 2023 open-weight release that kicked off the open LLM boom
Llama 2 Meta's 2023 open-weight release that jump-started the open LLM ecosystem
Llama 3 Meta's 2024 open-weight release that closed most of the gap to closed models
Llama Guard 2 Meta's earlier content-safety classifier model, preceding Llama Guard 3
Luma Dream Machine 1.0 Luma's first publicly available text/image-to-video model
Megatron-Turing NLG NVIDIA and Microsoft's 530B-parameter research model, among the largest dense LLMs of its time
Mistral Large 2 Mistral's 2024 flagship model, the generation preceding Mistral Large 3
MPT-30B MosaicML's open-weight commercial-use model, later folded into Databricks
mT5 Google's multilingual variant of T5, covering 101 languages
Muse Google's masked-transformer text-to-image model, faster than comparable diffusion models
MusicLM Google's early text-to-music generation model
Nomic Embed Text v1 Nomic's first fully open-source, open-weight, open-data long-context embedding model
NVIDIA GeForce GTX 980 Ti Flagship Maxwell-era consumer GPU
NVIDIA Tesla K80 Dual-GPU Kepler-era compute accelerator
NVIDIA Tesla M60 Dual-GPU Maxwell-era virtualization card
NVIDIA Titan Xp Prosumer flagship of the Pascal generation
o1-mini OpenAI's first compact reasoning model, tuned for coding and math
o1-preview Public preview of OpenAI's first reasoning model, ahead of the full o1 release
OLMo 7B AI2's first fully open model release, with open weights, data, and training code
OPT-175B Meta's 2022 open-weight model released to mirror GPT-3's scale
PaLM Google's 2022 540B-parameter model that set the stage for Gemini
PaLM 2 Google's 2023 large language model that powered the original Bard
PanGu-Alpha Huawei's first large-scale Chinese autoregressive language model
Parti Google's autoregressive text-to-image model, an alternative approach to diffusion
Phi-2 Microsoft's earlier 2.7B model demonstrating outsized small-model performance
Qwen-7B Alibaba's original Qwen release that started the model family
Qwen-Audio Alibaba's first audio-language model for speech and sound understanding
Qwen-VL Alibaba's first vision-language model in the Qwen family
Qwen1.5-110B Alibaba's largest dense model in the Qwen1.5 generation
Qwen2-72B Alibaba's second-generation flagship dense open-weight model
QwQ-32B-Preview Alibaba's first public preview of its QwQ reasoning line, ahead of the full QwQ-32B release
Recraft V2 Recraft's earlier image and vector generation model, preceding V3
RoBERTa Meta's robustly-optimized retraining of BERT that improved on it across benchmarks
Runway Gen-1 Runway's first video-to-video generation model
Runway Gen-2 Runway's first text-to-video model, preceding Gen-3 Alpha
Runway Gen-3 Alpha Runway's 2024 text-to-video generation model
Samsung Gauss Samsung's original in-house foundation model, ahead of Gauss 2
SantaCoder BigCode's small multilingual code model preceding StarCoder
SenseChat-3 SenseTime's third-generation SenseChat model, preceding SenseChat-5
SenseNova 5.0 SenseTime's flagship multimodal SenseNova model, ahead of the 5.5 update
Skywork-13B Kunlun Tech's open-weight bilingual foundation model
SmolLM Hugging Face's original family of small, fully open language models
Solar 10.7B Upstage's depth-upscaled open model that preceded the Solar Pro line
Sora OpenAI's original text-to-video model, notable for minute-long coherent generations at launch
Sparrow DeepMind's dialogue agent research model trained to be more helpful and less harmful
StableLM 3B Stability AI's early compact open-weight language model, preceding StableLM 2
StarCoder The original BigCode model that StarCoder2 later succeeded
Step-1 StepFun's earlier flagship model that preceded Step-2
T5 Google's text-to-text transfer transformer, an early unifying NLP framework
Tacotron 2 Google's neural text-to-speech model pairing a spectrogram predictor with a WaveNet vocoder
text-davinci-003 The last and most capable of OpenAI's original GPT-3.5 completion models
text-embedding-ada-002 OpenAI's long-serving default embedding model before the v3 family
Titan Text Amazon's original in-house foundation model line for Bedrock
Titan Text Express Amazon's mid-tier Titan model for general text generation, between Lite and the base Titan Text
Turing-NLG Microsoft's 17B-parameter model that was among the largest published language models of its time
UL2 Google's unified pretraining framework blending multiple denoising objectives
Veo Google's original Veo text-to-video model, previewed at I/O 2024 ahead of Veo 3
Vicuna-13B LMSYS's influential early open chat model fine-tuned from Llama on ShareGPT conversations
Vicuna-33B The largest Vicuna release, LMSYS's ShareGPT fine-tune of Llama
Voyage-2 Voyage AI's general-purpose embedding model, preceding Voyage-3
Wav2Vec 2.0 Meta's self-supervised speech representation model that underpins many ASR systems
WaveNet DeepMind's raw-audio generative model that set the foundation for modern neural TTS
WizardCoder Microsoft's code-specialised fine-tune built with Evol-Instruct data generation
XGen-7B Salesforce's open-weight long-context research model
XLNet A permutation-based autoregressive pretraining model that outperformed BERT on many tasks