New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
Large language models (LLMs)
Browse every major large language model in one place. This LLM list tracks frontier and open-source foundation models — GPT, Claude, Gemini, Llama, Mistral, Qwen and 400+ more — with parameter counts, context windows, benchmark scores and provider pricing.
Sandwich Transformer
🇺🇸 Allen Institute for AI
XLM-RoBERTa
🇺🇸 Facebook AI
Base LM + kNN LM + Continuous Cache
🇺🇸 Stanford University
BART-large
🇺🇸 Facebook AI
T5-11B
T5-3B
RMSNorm (Transformer-base)
🇺🇸 University of Edinburgh
M4-50B
DistilBERT
🇺🇸 Hugging Face
ALBERT
🇺🇸 Toyota Technological Institute at Chicago
Adaptive Inputs + LayerDrop
🇺🇸 Facebook AI Research
Megatron-BERT
🇺🇸 NVIDIA
Megatron-LM (1.2B)
🇺🇸 NVIDIA
Megatron-LM (2.5B)
🇺🇸 NVIDIA
Megatron-LM (355M)
🇺🇸 NVIDIA
Megatron-LM (8.3B)
🇺🇸 NVIDIA
Xiaoice
🇺🇸 Microsoft Research Asia
DEQ-Transformer (Medium, Adaptive Embedding)
🇺🇸 Carnegie Mellon University (CMU)
DEQ-TrellisNet (PTB)
🇺🇸 Carnegie Mellon University (CMU)
DEQ-TrellisNet (WT-103)
🇺🇸 Carnegie Mellon University (CMU)
AWD-LSTM+Behaviorial-Gating
🇺🇸 University of Southern California
LSTM-Medium+Behaviorial-Gating (PTB)
🇺🇸 University of Southern California
trRosetta
🇺🇸 Nankai University
TripletRes
🇺🇸 University of Michigan
RNN Baseline
🇺🇸 Massachusetts Institute of Technology (MIT)
R-Transformer
🇺🇸 Michigan State University
All-attention network + adaptive span
🇺🇸 Facebook AI Research
RoBERTa Base
RoBERTa Large
Tensorized Transformer (257M)
🇺🇸 Tianjin University
Tensorized Transformer (OBW)
🇺🇸 Tianjin University
Tensorized Transformer (PTB)
🇺🇸 Tianjin University
Tensorized Transformer (W103)
🇺🇸 Tianjin University
Tensorized Transformer (large PTB)
🇺🇸 Tianjin University
TAPE Transformer
🇺🇸 University of California (UC) Berkeley
SAGAN
🇺🇸 Rutgers University
Char-CNN-BiLSTM
🇺🇸 Capital One
4 layer QRNN + dynamic evaluation
🇺🇸 University of Texas at Austin
AWD-LSTM + MoS + Partial Shuffled
🇺🇸 University of Texas at Austin
AdvSoft + 4 layer QRNN + dynamic evaluation (WT103)
🇺🇸 University of Texas at Austin
Adversarial + AWD-LSTM-MoS + partial shuffled
🇺🇸 University of Texas at Austin
AWD-LSTM + Phrase Induction + finetuning (PTB)
🇺🇸 Massachusetts Institute of Technology (MIT)
Transformer-XL Large + Phrase Induction
🇺🇸 Massachusetts Institute of Technology (MIT)
VQ-VAE-2 (FFHQ)
VQ-VAE-2 (ImageNet)
XLNet
🇺🇸 Carnegie Mellon University (CMU)
DLRM-2020
🇺🇸 Facebook AI
Grover-Mega
🇺🇸 University of Washington
Flow++ (CIFAR10)
🇺🇸 University of California (UC) Berkeley
RaptorX-Contact
🇺🇸 Toyota Technological Institute at Chicago
Sparse Transformer (CIFAR10)
🇺🇸 OpenAI
Sparse Transformer (Enwik8)
🇺🇸 OpenAI
Sparse Transformer (ImageNet)
🇺🇸 OpenAI
BERT-Large-CAS (PTB+WT2+WT103)
🇺🇸 Amazon
BERT-Large-CAS (WT103)
🇺🇸 Amazon
BERT-Large-CAS (WT2)
🇺🇸 Amazon
SpecAugment
🇺🇸 Google Brain
Transformer-XL + RMS dynamic eval
🇺🇸 University of Edinburgh
MEGNet (crystal band gap model)
🇺🇸 University of California San Diego
MEGNet (crystal elasticity model)
🇺🇸 University of California San Diego
About Large language models (LLMs)
Large language models (LLMs) are the foundation of modern generative AI — general-purpose text models trained on vast corpora that can write, reason, summarise, translate and code. Choosing the best LLM is rarely about a single winner: the best AI model for one task may lag on another. Reasoning-heavy work rewards models that score well on benchmarks like MMLU, GPQA and AIME, while agentic and tool-use workloads care more about instruction following, function calling and long-context recall. When you compare LLMs, weigh raw capability against the practical constraints that decide cost and feasibility — context window, throughput, latency, licensing and price per million tokens. Open-source LLMs such as Llama, Qwen, Mistral and DeepSeek let you self-host and fine-tune, while proprietary frontier models from OpenAI, Anthropic and Google often lead on raw quality. Use our benchmarks to see where each model ranks, and put two candidates side by side with compare before you commit to a provider.
Frequently asked questions
What is the best LLM right now?
There is no single best LLM — it depends on the task. Frontier proprietary models from OpenAI, Anthropic and Google tend to lead on reasoning benchmarks, while open-source LLMs like Llama, Qwen and DeepSeek are best when you need to self-host or fine-tune. Compare candidates on the benchmarks page for your specific workload.
What is the best open source LLM?
The strongest open-source and open-weights LLMs at any given time typically come from the Llama, Qwen, Mistral and DeepSeek families. They can be downloaded, self-hosted and fine-tuned, and the top ones rival proprietary models on many benchmarks. Filter the list above by availability to see current open-weights options.
How do I compare two LLMs?
Use the compare tool to put two models side by side on parameters, context window, availability and benchmark scores, then check the benchmarks page for task-specific rankings such as MMLU, GPQA and coding scores.