Chatting

Models 162

• Gen 7
  Instella MoE 16B A3B DPO

  By AMD

  Instella-MoE-16B-A3B-DPO is the direct preference optimization checkpoint of AMD's Instella-MoE model, a fully open Mixture-of-Experts language model with 16 billion total parameters and 2.8 billion active parameters per token. Built on the SFT checkpoint, it is tuned on contrastive preference data to improve response quality and alignment, serving as an input to the final reinforcement-learning stage.

  💬Chatting💻Coding🔍Advanced reasoning
  NewText

  Released 4d ago
• Gen 7
  Instella MoE 16B A3B SFT

  By AMD

  Instella-MoE-16B-A3B-SFT is the supervised fine-tuning checkpoint of AMD's Instella-MoE model, a fully open Mixture-of-Experts language model with 16 billion total parameters and 2.8 billion active parameters per token. Built on the long-context base checkpoint, it is fine-tuned to follow instructions and perform chain-of-thought reasoning, serving as the input to the later DPO and reinforcement learning stages.

  💬Chatting💻Coding🔍Advanced reasoning
  NewText

  Released 4d ago
• Gen 7
  Instella MoE 16B A3B Base

  By AMD

  Instella-MoE-16B-A3B-Base is the long-context checkpoint AMD designates as the final base model in its Instella-MoE family, a fully open Mixture-of-Experts language model with 16 billion total parameters and 2.8 billion active parameters per token. It extends the mid-trained checkpoint with long-context training so it can process and reason over longer sequences, and serves as the foundation for the SFT, DPO, and RL post-trained checkpoints.

  💬Chatting💻Coding
  NewText

  Released 4d ago
• Gen 7
  Instella MoE 16B A3B Midtrain

  By AMD

  Instella-MoE-16B-A3B-Midtrain is the mid-training checkpoint of AMD's Instella-MoE model, a fully open Mixture-of-Experts language model with 16 billion total parameters and 2.8 billion active parameters per token. Built on the pre-trained checkpoint, it is further trained on high-quality data mixtures to refine core language capabilities on AMD Instinct MI300X and MI325X GPUs, ahead of long-context extension and instruction tuning stages.

  💬Chatting💻Coding
  NewText

  Released 4d ago
• Gen 7
  Instella MoE 16B A3B Pretrain

  By AMD

  Instella-MoE-16B-A3B-Pretrain is the base pre-training checkpoint of AMD's Instella-MoE model, a fully open Mixture-of-Experts language model with 16 billion total parameters and 2.8 billion active parameters per token. Trained from scratch on AMD Instinct MI300X and MI325X GPUs, it uses Gated Multi-head Latent Attention and FarSkip-Collective communication for efficient large-scale training, serving as the starting checkpoint for later mid-training and fine-tuning stages.

  💬Chatting💻Coding
  NewText

  Released 4d ago
• Gen 7
  Instella MoE 16B A3B Think

  By AMD

  Instella-MoE-16B-A3B-Think is the final reinforcement-learning tuned checkpoint of AMD's Instella-MoE model, a fully open Mixture-of-Experts language model with 16 billion total parameters and 2.8 billion active parameters per token. Refined with RL on top of the DPO checkpoint, it is the strongest reasoning and instruction-following model in the release, generating extended chain-of-thought responses up to 32768 tokens.

  💬Chatting📝Content creation💻Coding🔍Advanced reasoning
  NewText

  Released 4d ago
• Gen 3 Claude
  Claude Opus 5

  By Anthropic

  Claude Opus 5 is Anthropic's flagship Opus tier large language model, built for long running agentic work, coding, and professional knowledge tasks. It approaches the intelligence of Anthropic's top tier model at roughly half the cost, with stronger results on software engineering, computer use, and scientific research benchmarks than its predecessor, Opus 4.8.

  💬Chatting🤖Task automation🔎Problem solving💻Coding🤖Agents🔬Scientific research
  NewMultimodal

  Released 4d ago
• Gen 7 Claude
  Opus 5

  By Anthropic

  Claude Opus 5 is Anthropic's new Opus flagship, released July 24, 2026. It's noticeably stronger at coding, research, and long, multi step tasks than Opus 4.8, at the same price, and comes close to Anthropic's top tier model (Fable 5) at half the cost.

  🚀Productivity💻Coding💬Chatting
  NewText

  Released 4d ago
• Gen 7 Ling
  Ling 3.0 Flash

  By InclusionAI

  Ling-3.0-flash is a 124B-parameter Mixture-of-Experts language model with about 5.1B parameters activated per token. It is built for token efficiency and production-scale agentic inference, supporting a 262K token context window for text-based conversation, coding, and tool-driven workflows.

  💬Chatting🤖Task automation🔎Problem solving💻Coding📝Writing assistance
  NewText

  Released 5d ago
• Gen 7
  Nanbeige4.2-3B

  By Nanbeige LLM Lab

  Nanbeige4.2-3B is a compact 3B-parameter agentic language model using a Looped Transformer architecture that reuses transformer layers to boost capacity without adding parameters. It combines strong tool-use and coding-agent performance with broad reasoning across math, science, and general agent benchmarks, supporting configurable thinking and tool-calling modes for multi-turn agentic workflows.

  💬Chatting💻Coding🤖Agents🔬Scientific research🔢Math
  NewText

  Released 6d ago
• Gen 7 Qwen
  Macaron V1 Tall

  By Macaron

  Macaron-V1-Tall is a 50 billion parameter agent model post-trained from Qwen 3.6, using the same Mixture-of-LoRA architecture as its larger sibling but sized for local deployment. Handles chat, agentic tool use, coding, and generative UI rendering with a smaller compute footprint.

  💬Chatting💻Coding🤖Agents
  NewText

  Released 7d ago
• Gen 7 GLM
  Macaron V1 Venti

  By Macaron

  Macaron-V1-Venti is a 748 billion parameter agent model post-trained on GLM-5.2, combining a frozen base with four Mixture-of-LoRA specialists for chat, agentic tool use, coding, and generative UI rendering. Built for long-horizon personal-life and coding tasks with strong tool-use and reasoning performance versus other frontier general-purpose models.

  💬Chatting🤖Task automation💻Coding🤖Agents💼Personal assistant
  NewText

  Released 7d ago
• Gen 3 Gemini
  Gemini 3.5 Flash Lite

  By Google

  Gemini 3.5 Flash Lite is Google's fastest, most cost efficient model in the Gemini 3 series. It is a natively multimodal reasoning model built on Gemini 3.1 Flash Lite, optimized for high volume, latency sensitive tasks like translation, classification, and document processing, while also supporting agentic workflows.

  💬Chatting💻Coding🔎Search🌍Translations📄Document processing
  NewMultimodal

  Released 7d ago
• Gen 3 Gemini
  Gemini 3.6 Flash

  By Google

  Gemini 3.6 Flash is Google's workhorse model in the Gemini 3 series, a natively multimodal reasoning model built on Gemini 3.5 Flash. It improves coding, knowledge work, and multimodal performance while using fewer output tokens and running at a lower cost per task than its predecessor.

  💬Chatting📊Data analysis💻Coding📷Images🎥Videos
  NewMultimodal

  Released 7d ago
• Gen 3
  Kimi K3

  By Moonshot AI

  Kimi K3 is a 2.8 trillion parameter sparse Mixture-of-Experts language model with native vision understanding and a context window of up to 1 million tokens. It uses Kimi Delta Attention and Attention Residuals to improve long-context efficiency and depth scaling, with an always-on reasoning mode for coding, knowledge work, and agentic tool use.

  💬Chatting💻Coding🤖Agents🔍Research📄Document analysis🔍Image analysis
  NewMultimodal

  Released 12d ago
• Gen 3
  Inkling

  By Thinking Machines Lab

  Inkling is an open-weights Mixture-of-Experts transformer from Thinking Machines Lab with 975B total parameters (41B active) and native multimodal support for text, image, and audio input with text output. It supports up to a 1M-token context window and controllable reasoning effort, letting developers trade off cost and latency. Trained on 45 trillion tokens, it targets broad, balanced performance across reasoning, agentic coding, tool use, and multimodal tasks.

  💬Chatting💻Coding🗒Transcription🔍Image analysis
  NewMultimodal

  Released 13d ago
• Gen 3
  Bonsai 27B AWQ 4bit

  By PrismML

  Bonsai 27B AWQ 4bit is a 4-bit AWQ quantized release of Prism ML's Bonsai 27B, a 27 billion parameter multimodal model that accepts image and text input. It supports multi-step reasoning, tool calling, and agentic workflows, and is built for efficient deployment on inference engines like vLLM and SGLang.

  💬Chatting📷Images🔍Image analysis
  NewMultimodal

  Released 13d ago
• Gen 3 Qwen
  Ternary Bonsai 27B AWQ 4bit

  By PrismML

  Ternary Bonsai 27B AWQ 4bit is a 4 bit AWQ quantized build of PrismML's Ternary Bonsai 27B, a multimodal model based on Qwen3.6 27B. It accepts text and images, supports a 262K token context window, and is designed for reasoning, coding, and agentic tool use workflows served efficiently via vLLM or SGLang.

  💬Chatting💡Coding help🔍Image analysis
  NewMultimodal

  Released 14d ago
• Gen 3 Qwen
  Ternary Bonsai 27B Unpacked

  By PrismML

  Ternary Bonsai 27B is a 27B parameter multimodal language model from PrismML, built on a hybrid linear and full attention (Qwen3.5 based) architecture with a 262,144 token context window. This repository provides the full size FP16 safetensors weights, an alternative to the natively packed 2 bit ternary format for use with standard model-serving tooling.

  💬Chatting📷Images🤖Agents
  NewMultimodal

  Released 14d ago
• Gen 3 Qwen
  Bonsai 27B

  By PrismML

  Bonsai 27B is a 27.8B-parameter multimodal language model built on Qwen3.6 27B, quantized end-to-end to 1-bit or 1.58-bit ternary weights across embeddings, attention, MLPs, and the LM head. It accepts text and image input with a 262,144-token context window, and supports reasoning, tool use, and coding, while running locally on phones, laptops, and edge GPUs.

  💬Chatting💻Coding🔢Math
  NewMultimodal

  Released 14d ago
• Gen 3
  Muse Spark 1.1

  By Meta Platforms

  Muse Spark 1.1 is a multimodal reasoning model built for agentic tasks, with strengths in tool use, computer use, coding, and multimodal understanding. It manages a 1 million token context window and orchestrates multi-agent systems for complex, long-running projects.

  💬Chatting🎥Video analysis💻Coding🤖Agents🔍Image analysis
  NewMultimodal

  Released 19d ago
• Gen 7 Hunyuan
  Hy3 295B

  By Tencent

  Hy3 is a 295B-parameter Mixture-of-Experts language model with 21B active parameters, developed by the Tencent Hy team. It offers strong reasoning, coding, and agentic capabilities with a 256K token context window, improved tool-call reliability, and reduced hallucination compared to its preview version.

  💬Chatting💻Coding🤖Agents🚀Productivity🔍Advanced reasoning
  NewText

  Released 22d ago
• Gen 7 Hunyuan
  Hy3

  By Tencent

  Hy3 is a 295 billion parameter Mixture of Experts language model from Tencent with 21 billion active parameters and a 256K token context window. It supports fast and deep reasoning modes, tool calling, and multi turn dialogue, with improved reliability, reduced hallucination, and stronger agentic and coding performance over its preview version.

  💬Chatting📝Content creation💻Coding🤖Agents🤔Logical reasoning
  NewText

  Released 22d ago
• Gen 3 Qwen
  Ternary Bonsai 27B

  By PrismML

  A 27B parameter language model compressed to ternary (1.71 bit) weights, retaining 95% of full precision intelligence while shrinking deployed size to about 7.2 GB. Runs full 27B class reasoning, math, and coding on a standard laptop or single GPU with a 262K token context window.

  💬Chatting💻Coding🔢Math
  NewMultimodal

  Released 27d ago
• Gen 3 Qwen
  Ternary Bonsai 27B GGUF

  By PrismML

  Ternary Bonsai 27B GGUF is a 27B parameter reasoning language model compressed to true 2 bit ternary weights (1.71 bits per weight), derived from a Qwen3.6-27B hybrid attention backbone. It runs in a roughly 7.2GB footprint on laptops, single GPUs, and CPUs via llama.cpp, while retaining strong math, coding, tool use, and long context (262K token) reasoning performance.

  💬Chatting💻Coding🔢Math
  NewMultimodal

  Released 27d ago
• Gen 3 Qwen
  Bonsai 27B mlx 1bit

  By PrismML

  1-bit quantized version of the 27B-parameter Qwen3.6-27B language model, using binary g128 weight representation (1.125 bits per weight) for a ~3.9GB deployed footprint, about 14.2x smaller than FP16. Supports a 262K token context, retains around 89.5% of FP16 accuracy across reasoning, math, coding, and tool-use benchmarks, and runs on-device on phones, laptops, and GPUs via Apple MLX and CUDA.

  💬Chatting💻Coding🔢Math
  NewMultimodal

  Released 27d ago
• Gen 3 Claude
  Claude Sonnet 5

  By Anthropic

  Claude Sonnet 5 is Anthropic's large language model, the most agentic Sonnet released so far. It plans autonomously, uses tools such as browsers and terminals, and sustains long multi step coding and computer use tasks. Its agentic performance on reasoning, tool use, and coding approaches Opus 4.8 while costing less, and it supports selectable effort levels for balancing speed, cost, and capability.

  💬Chatting🤖Task automation🔎Problem solving💻Coding📝Writing assistance
  NewMultimodal

  Released 28d ago
• Gen 7 Apertus
  Apertus v1.1 4B Instruct

  By Swiss AI Initiative

  A 4B parameter instruction-tuned language model created via pre-training distillation from the Apertus-8B teacher model. Uses a dense transformer decoder with grouped-query attention and xIELU activations, trained on 1.7T tokens. Natively supports 1811 languages with quantized checkpoints available for mobile and edge deployment. Apache-2.0 licensed.

  💬Chatting💻Coding📝Writing🌍Translations
  NewText

  Released 29d ago
• Closed betaGen 7 GPT
  GPT 5.6 Terra

  By OpenAI

  GPT 5.6 Terra is a balanced model in OpenAI's GPT 5.6 preview series. It offers performance competitive with GPT 5.5 while costing about half as much. Terra shows improved coding and cybersecurity capability that scales with reasoning effort, and is paired with a layered safety and misuse safeguard stack during its limited preview release.

  💬Chatting🐞Debugging💻Coding🔒Cybersecurity🔒AI security
  NewText

  Released 1mo ago
• Closed betaGen 7 GPT
  GPT 5.6 Luna

  By OpenAI

  GPT 5.6 Luna is the fast and affordable tier of OpenAI's GPT 5.6 preview series. It is the lowest cost model in the family while still delivering strong capability, including cybersecurity gains that scale with reasoning effort. Luna is released as part of a limited, safeguard paired preview alongside the Sol and Terra models.

  💬Chatting🐞Debugging💻Coding🔒Cybersecurity🔒AI security
  NewText

  Released 1mo ago
• Closed betaGen 7 GPT
  GPT 5.6 Sol

  By OpenAI

  GPT-5.6 Sol is OpenAI's flagship model in the GPT-5.6 series. It achieves state-of-the-art performance on coding (Terminal-Bench 2.1), biology (GeneBench v1), and cybersecurity (ExploitBench) benchmarks. Features a new max reasoning effort setting and ultra mode for multi-agent parallelism. Priced at $5/$30 per 1M input/output tokens, currently in limited preview.

  💬Chatting📊Data analysis🔎Problem solving💻Coding🔍Research🔬Scientific research
  NewText

  Released 1mo ago
• Gen 3 GLM
  GLM 5.2

  By Z.ai

  GLM-5.2 is Z.AI’s high-capability model for complex reasoning, large-scale engineering, agentic coding, and difficult long-horizon development tasks.

  💬Chatting🔎Problem solving💻Coding🤖Agents💻Conversational coding💻Vibe coding
  NewMultimodal

  Released 1mo ago
• Gen 3
  DiffusionGemma

  By Google

  DiffusionGemma is Google’s open experimental 26B MoE text-diffusion model for faster local text generation, parallel editing, code infilling, and non-linear text workflows.

  💬Chatting🔄Language model optimization🧠AI inference
  NewMultimodal

  Released 1mo ago
• Gen 3 Claude
  Claude Mythos 5

  By Anthropic

  Claude Mythos 5 is Anthropic’s restricted-access version of the same underlying model as Claude Fable 5, designed for trusted cyberdefense, infrastructure security, and selected high-impact research use.

  💬Chatting🔒Cybersecurity🤖Agents🔬Scientific research🧬Protein design🧬Life sciences
  NewMultimodal

  Released 1mo ago
• Gen 7 LFM
  LFM 2.5 1.2B JP 202606

  By Liquid AI

  LFM2.5-1.2B-JP-202606 is Liquid AI’s 1.17B Japanese-focused text model for chat, instruction following, math, code, tool use, and bilingual English-Japanese assistant workflows.

  💬Chatting🈯️Japanese content generation
  NewText

  Released 1mo ago
• Gen 7
  Aion 1.0 Instruct

  By Microsoft

  Aion 1.0 Instruct is Microsoft’s on-device small language model for local text intelligence such as summarization, rewriting, intent handling, and accessibility features on Windows.

  💬Chatting🔄Text rewriting🔍Intent recognition
  NewText

  Released 1mo ago
• Gen 7Nemotron
  Nemotron 3 Ultra 550B A55B NVFP4

  By NVIDIA

  NVIDIA Nemotron-3-Ultra-550B-A55B-NVFP4 is NVIDIA’s open 550B total, 55B active frontier reasoning LLM for agentic workflows, long-context analysis, tool use, code, math, and science.

  💬Chatting🔎Problem solving🤖Agents
  NewText

  Released 1mo ago
• Gen 3 MAI
  MAI Thinking 1

  By Microsoft

  MAI-Thinking-1 is Microsoft AI’s flagship reasoning model for complex reasoning, software engineering, math, and high-difficulty problem solving.

  💬Chatting🔎Problem solving🤔Logical reasoning💻Conversational coding🤔Stem problem solving
  NewMultimodal

  Released 1mo ago
• Gen 3
  Agnes 2.0 Flash

  By Agnes AI

  Agnes 2.0 Flash is a fast, cost efficient language model built for high frequency production use. It handles text generation, multi turn conversation, tool calling, coding assistance, agent workflows, and image understanding via image URLs, with an optional Thinking mode for deeper step by step reasoning on complex tasks.

  💬Chatting🤖Task automation💻Coding🤖Agents🔍Image analysis
  NewMultimodal

  Released 1mo ago
• Gen 3 MiniMax
  MiniMax M3

  By MiniMax

  MiniMax M3 is MiniMax’s open-weight multimodal model for agentic coding, tool use, long-context tasks, and native text-visual reasoning.

  💬Chatting🤖Agents🔍Image interpretation💻Conversational coding💻Vibe coding
  NewMultimodal

  Released 1mo ago
• Gen 3
  Qwen3.7-Plus

  By Alibaba

  Qwen3.7-Plus is Alibaba Qwen’s multimodal agent model that unifies vision and language for agentic vision-language workflows.

  💬Chatting🤖Agents🔍Image interpretation💻Conversational coding
  NewMultimodal

  Released 1mo ago
• Gen 3 Step
  Step 3.7 Flash

  By StepFun

  StepFun's flagship multimodal reasoning model, a 198B total parameter sparse MoE with 11B active parameters. Natively understands images and video, supports adjustable reasoning effort, and is optimized for agent and coding workflows with reliable tool calling.

  💬Chatting🤖Task automation🎥Video analysis💻Coding🤖Agents🔍Image analysis
  NewMultimodal

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

  By Swiss AI Initiative

  Apertus-v1.1-1.5B-Instruct is a 1.5B parameter instruction-tuned language model built on a dense transformer decoder with grouped-query attention and xIELU activation. Created via pre-training distillation from an 8B teacher model trained on 1.7T tokens. Natively supports 1811 languages and optimized for mobile and edge deployments. Apache 2.0.

  💬Chatting🌐Text translation📝Writing📝Text editing
  NewText

  Released 2mo ago
• Gen 3
  MiniCPM5 1B

  By OpenBMB

  MiniCPM5-1B is OpenBMB’s open-source 1B causal language model for on-device local assistants, coding agents, tool use, and compact reasoning.

  💬Chatting➗Math problems🤔Logical reasoning💻Conversational coding
  NewMultimodal

  Released 2mo ago
• Gen 3 Qwen
  Qwen 3.7 Max

  By Alibaba

  Qwen3-Max is Alibaba Cloud’s closed-source trillion-parameter flagship LLM for coding, reasoning, enterprise agentic AI, and tool-use workflows.

  💬Chatting🤖Task automation🤔Logical reasoning💻Conversational coding
  NewMultimodal

  Released 2mo ago
• Gen 3 Command
  Command A+ W4A4

  By Cohere

  Command A+ 05-2026 W4A4 is Cohere’s open-source quantized vision-language reasoning model for agentic, multilingual, tool-use, and enterprise tasks.

  💬Chatting🔍Image interpretation💻Conversational coding
  NewMultimodal

  Released 2mo ago
• Gen 3Gemini
  Gemini 3.5 Flash

  By Google DeepMind

  Gemini 3.5 Flash is Google’s natively multimodal reasoning model focused on frontier performance with controllable quality, cost, and latency.

  💬Chatting
  NewMultimodal

  Released 2mo ago
• Gen 7 Nemotron
  Nemotron Labs Diffusion 14B

  By NVIDIA

  Nemotron-Labs-Diffusion-14B is NVIDIA’s open 14B text-generation language model that supports autoregressive decoding, diffusion-style parallel decoding, and self-speculative decoding.

  💬Chatting🌐Text translation💻Conversational coding
  NewText

  Released 2mo ago
• Gen 3
  Ring 2.6 1T

  By AntGroup

  Ring-2.6-1T is inclusionAI’s open-source 1T-parameter text-generation reasoning model for agent workflows, coding, scientific analysis, and enterprise automation.

  💬Chatting🤖Task automation🔄Workflows💻Conversational coding💻Vibe coding
  NewMultimodal

  Released 2mo ago
• Gen 3
  MiniCPM V4.6.

  By OpenBMB

  MiniCPM-V-4.6 is OpenBMB’s open-source lightweight multimodal model for efficient image, multi-image, and video understanding on mobile and edge devices.

  💬Chatting🎥Video analysis🔍Image interpretation📜OCR
  NewMultimodal

  Released 2mo ago