Large Language Models

Models 20

• Gen 7 LongCat
  LongCat 2.0

  By Meituan

  LongCat-2.0 is an open-source MoE language model with 1.6 trillion total parameters (~48B activated per token) and a 1M-token context window. It introduces LongCat Sparse Attention for efficient long-context processing and was pretrained on 35+ trillion tokens using AI ASIC hardware. Strong focus on coding, agentic, and reasoning tasks.

  📚Large Language Models💻Coding🤖Agents🔢Math
  NewText

  Released 25d ago
• Gen 7 Apertus
  Apertus v1.1 4B

  By Swiss AI Initiative

  Apertus-v1.1-4B is a 4B parameter open-source base language model created via pre-training distillation from the Apertus-8B teacher model. It uses a dense transformer with grouped-query attention and xIELU activations, natively supports 1,811 languages, and is trained on 1.7T high-quality tokens. Designed for constrained hardware and further fine-tuning.

  📚Large Language Models🔍Research🌐Multilingual communication🧠Model training
  NewText

  Released 26d ago
• Gen 3 Ornith
  Ornith 1.0 397B

  By DeepReinforce

  Ornith-1.0-397B is a 397B MoE open-source reasoning model for agentic coding, post-trained on Qwen 3.5 MoE via a self-improving RL framework that jointly learns task solutions and the scaffolds guiding them. Achieves 82.4 on SWE-Bench Verified and 77.5 on Terminal-Bench 2.1. MIT licensed with tool-calling and 256K context support.

  🔍Code analysis🐞Debugging📚Large Language Models🤖Agents💻Vibe coding
  NewMultimodal

  Released 28d ago
• Gen 7 LFM
  LFM2.5 230M

  By Liquid AI

  LFM2.5-230M is a 230M-parameter lightweight language model built on the LFM2 architecture, designed for fast inference across edge and cloud environments. Pre-trained on 19T tokens with a 32K context window, it targets tool use, data extraction, and agentic workflows. Runs on CPUs including Raspberry Pi 5 and mobile SoCs. Not recommended for math, code generation, or creative writing.

  🤖Task automation📚Large Language Models🤖Agents🤖Robotics
  NewText

  Released 1mo ago
• Gen 7 Apertus
  Apertus v1.1 1.5B

  By Swiss AI Initiative

  Apertus-v1.1-1.5B is a 1.5 billion parameter base language model trained via distillation from an 8B teacher model. It supports 1,811 languages, uses a dense transformer architecture with grouped-query attention and xIELU activation, and has a 32,768 token context window. Designed for text generation and multilingual applications.

  📚Large Language Models🌍Translations🔄Language model optimization
  NewText

  Released 1mo ago
• Gen 3
  ZAYA1 8B

  By Zyphra AI

  ZAYA1-8B is Zyphra’s small mixture-of-experts reasoning model, built for long-form reasoning, mathematics, and coding. It has 8.4B total parameters with 760M active parameters, is post-trained for reasoning, and is positioned as highly intelligence-efficient enough for both strong test-time-compute use and local on-device deployment.

  📚Large Language Models🔎Problem solving💻Coding🔢Math
  NewMultimodal

  Released 2mo ago
• Gen 3
  SubQ 1M Preview

  By Subquadratic

  SubQ 1M-Preview is Subquadratic’s first large language model and the first fully subquadratic LLM, built so compute grows linearly with context length rather than quadratically. It is designed for massive-context reasoning, coding, search, and full-corpus processing, making million-token windows more practical, faster, and cheaper than standard transformer-based approaches

  📚Large Language Models🧠AI inference🦾Ai research
  NewMultimodal

  Released 2mo ago
• Gen 7
  nanowhale

  By Hugging Face

  nanowhale is a miniature DeepSeek-V4-style language model project, built to reproduce the core DeepSeek-V4 architecture at about 110M parameters. It is designed as an educational and research-scale model, with both a pretrained base version and an SFT chat version, rather than as a production frontier assistant.

  📚Large Language Models🧠Model training🦾Ai research
  NewText

  Released 2mo ago
• Gen 3 DeepSeek
  DeepSeek V4 Flash

  By DeepSeek

  DeepSeek-V4-Flash is DeepSeek’s faster, smaller, and more economical V4 model for efficient large-scale use. It has 284B total parameters with 13B active parameters, supports a 1M token context window, and is positioned as a high-speed, cost-effective model whose reasoning comes close to V4-Pro.

  📚Large Language Models💬Chatting💻Conversational coding🧠AI inference
  Multimodal

  Released 3mo ago
• Gen 3 DeepSeek
  Deepseek V4 Pro

  By DeepSeek

  DeepSeek-V4-Pro is DeepSeek’s new flagship open-source model for top-end reasoning, coding, world knowledge, and agentic work. It uses a MoE architecture with 1.6T total parameters and 49B active parameters, supports a 1M token context window, and is positioned as rivaling leading closed-source models while leading current open models on several major capabilities.

  📚Large Language Models💬Chatting💻Conversational coding🧠AI inference
  Multimodal

  Released 3mo ago
• Gen 3 Ling
  Ling 2.6 1T

  By AntGroup

  Ling-2.6-1T is inclusionAI’s trillion-parameter flagship instruct model for real-world agents. It is positioned for fast execution, high efficiency at scale, advanced coding, complex reasoning, and large-scale agent workflows, using a “fast thinking” approach to cut cost while keeping top-tier performance.

  📚Large Language Models💬Chatting💻Conversational coding
  Multimodal

  Released 3mo ago
• Gen 7
  F pre 8kv 8k 14k SWA

  By Ai2

  F_pre_8kv_8k_14k_SWA is an AllenAI OlmPool architectural-variation model for studying long-context extension. It is a 7B to 8B class language model trained to 150B tokens total, including 140B pretraining and 10B context extension, and is meant for research on architecture choices rather than polished instruction-following use.

  📚Large Language Models🔄Language model optimization
  Text

  Released 3mo ago
• Gen 7
  Parcae

  By Sandy Research

  Parcae is a Sandy Research model family described as "a family of stable looped models." The public collection currently lists four sizes: 140M, 370M, 770M, and 1.3B.

  📚Large Language Models🔨LLM development🎓LLM training
  Text

  Released 3mo ago
• Gen 7
  LFM2.5 350M

  By Liquid AI

  LFM2.5-350M is Liquid AI’s compact 350M-parameter language model optimized for fast inference, tool use, data extraction, and structured outputs. It is designed to run across cloud GPUs, CPUs, and edge devices, and is positioned for lightweight agentic workflows and large-scale data processing rather than math, coding, or creative writing.

  📚Large Language Models🔍Data extraction🔄Language model optimization🧠AI inference
  Text

  Released 3mo ago
• Gen 3 MiMo
  Xiaomi MiMo V2 Omni

  By Xiaomi

  MiMo-V2-Omni is an omni foundation model that unifies multimodal understanding with agentic capability, built to see, hear, and act.

  📚Large Language Models🎥Video summaries📚Audio summaries🔍Image interpretation
  Multimodal

  Released 4mo ago
• Gen 3 MiMo
  Xiaomi MiMo V2 Pro

  By Xiaomi

  MiMo-V2-Pro is Xiaomi’s flagship foundation model built for real-world agent workloads, designed to act as the “brain” of agent systems that orchestrate complex workflows and tool use.

  📚Large Language Models🤖Agents🔄Workflow optimization💻Conversational coding
  Text

  Released 4mo ago
• Gen 7 Nemotron
  Nemotron 3 Super

  By NVIDIA

  Nemotron 3 Super is NVIDIA’s open Mixture-of-Experts hybrid Mamba-Transformer model designed for high-throughput agentic workloads and long-context reasoning.

  🤖Agents📚Large Language Models🔍Advanced reasoning💻Conversational coding
  Text

  Released 4mo ago
• Gen 7
  Sequential Hidden Decoding 8B n8

  By Tencent

  Sequential-Hidden-Decoding-8B-n8 is Tencent’s base language model variant that extends effective sequence length by 8x using multiple embedding passes while keeping the same Transformer backbone. Built on Qwen3-8B-Base, it is aimed at benchmarking, text completion, and fine-tuning research, with hidden latent steps acting like implicit reasoning before each predicted token.

  📚Large Language Models🤖Ai research assistance🔄Language model optimization🧠AI inference
  Text

  Released 4mo ago
• Gen 3
  Covenant 72B

  By Covenant AI

  Covenant-72B is a permissionless collaboratively trained 72B-parameter language model trained from scratch via decentralized coordination on Bittensor infrastructure.

  📚Large Language Models💬Chatting🔍Advanced reasoning
  Text

  Released 4mo ago
• Gen 3 GPT
  GPT 5.3

  By OpenAI

  GPT-5.3 is an updated general-purpose chat model optimized for smoother everyday conversations, more accurate answers, and better contextual understanding. It is positioned as a broadly useful assistant model that reduces unnecessary refusals, excessive caveats, and overly cautious phrasing, while keeping strong performance for common conversational and help-oriented tasks.

  📚Large Language Models💬Chatting🎙️Conversation enhancement🔍Conversational queries
  Multimodal

  Released 4mo ago