Chatting

Models 80

• Gen 4MiniMax
  MiniMax M2.7

  By MiniMax

  MiniMax M2.7 is MiniMax’s new text model release positioned around “self-evolution,” aimed at higher performance and value for complex tasks.

  💬Chatting💻Software engineering guidance💻Conversational coding🗣️Multi-agent conversations
  NewText

  Released 9d ago
• Gen 3GPT
  GPT 5.4 Mini

  By OpenAI

  GPT-5.4 mini is a fast, efficient GPT-5.4-family model optimized for high-volume coding and agent workloads, while keeping strong reasoning, multimodal understanding, and tool use.

  💬Chatting🎯Code autocompletion🔍Image interpretation💻Conversational coding
  NewMultimodal

  Released 10d ago
• Gen 7
  Nemotron Cascade 2 30B A3B

  By NVIDIA

  Nemotron-Cascade 2 is an open 30B Mixture-of-Experts model (3B activated) trained with Cascade RL and multi-domain on-policy distillation for strong reasoning and agentic capabilitie

  💬Chatting🔍Advanced reasoning💻Conversational coding⌨️Competitive programming coaching
  NewText

  Released 11d ago
• Gen 3Mistral
  Mistral Small 4

  By Mistral AI

  Mistral Small 4 is an open hybrid model that unifies instruct, reasoning, and coding in a single multimodal model with a 256k context window.

  💬Chatting🔍Advanced reasoning🔍Image interpretation🔢Math💻Conversational coding
  NewMultimodal

  Released 11d ago
• Gen 7
  GLM 5 Turbo

  By Z.ai

  GLM-5-Turbo is a foundation language model optimized for OpenClaw-style agent workflows, emphasizing reliable tool invocation, command following, scheduled and persistent tasks, and long-chain execution.

  🤖Agents🤖Task automation💬Chatting
  NewText

  Released 12d 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
  NewText

  Released 18d ago
• Gen 3GPT
  GPT 5.4 Pro

  By OpenAI

  GPT-5.4 Pro is the maximum-performance variant of GPT-5.4 for people who want the strongest results on complex tasks in ChatGPT and the API.

  💬Chatting🤖Agents🔍Advanced reasoning🧩Complex problem solving🔍Web research summarization💻Conversational coding
  NewMultimodal

  Released 22d ago
• Gen 3GPT
  GPT 5.4 Thinking

  By OpenAI

  GPT-5.4 Thinking is the ChatGPT-facing GPT-5.4 model mode tuned for professional work, emphasizing stronger reasoning, coding, and agentic workflows with fewer back-and-forth turns.

  💬Chatting🤖Agents🔍Advanced reasoning🧮Advanced math tutoring🧩Complex problem solving💻Conversational coding
  NewMultimodal

  Released 22d ago
• Gen 3GPT
  GPT 5.3 Instant

  By OpenAI

  GPT-5.3 Instant is an update to ChatGPT’s most-used everyday model, tuned for smoother, more consistently helpful conversations and stronger web-assisted answers.

  💬Chatting🔍Web search summaries💻Conversational coding🔍Comprehensive q&a
  NewMultimodal

  Released 24d ago
• Gen 3
  Qwen3.5 9B

  By Alibaba

  Qwen3.5-9B is a larger dense vision-language causal model with a vision encoder, targeting stronger capability for multimodal reasoning and agentic use

  💬Chatting🌐Text translation🔍Image interpretation💻Conversational coding
  NewMultimodal

  Released 25d ago
• Gen 3
  Qwen3.5 4B

  By Alibaba

  Qwen3.5-4B is a mid-size vision-language causal model with a vision encoder, designed for multimodal reasoning, coding, and agent workflows with very long context.

  💬Chatting🌐Text translation🔍Image interpretation💻Conversational coding
  NewMultimodal

  Released 25d ago
• Gen 3
  Qwen3.5 2B

  By Alibaba

  Qwen3.5-2B is a small vision-language causal model with a vision encoder, aimed at strong multimodal capability with efficient compute.

  💬Chatting🌐Text translation🔍Image interpretation💻Conversational coding
  NewMultimodal

  Released 25d ago
• Gen 3
  Qwen3.5 0.8B

  By Alibaba

  Qwen3.5-0.8B is a compact vision-language causal model with a vision encoder, built for multimodal understanding and agentic tool use at small scale.

  💬Chatting🌐Text translation🔍Image interpretation💻Conversational coding
  NewMultimodal

  Released 25d ago
• Gen 3
  Qwen 3.5 Flash

  By Alibaba

  Qwen3.5-Flash is the speed and cost optimized Qwen3.5 variant designed for high-throughput chat and multimodal prompting with a very long context window.

  💬Chatting📷Images🎥Videos
  NewMultimodal

  Released 1mo ago
• Gen 3Qwen
  Qwen3.5 397B A17B

  By Alibaba

  Qwen3.5-397B-A17B is a native vision-language model from Alibaba's Qwen team with 397B parameters (17B active per token). In the Qwen3.5-Plus hosted variant it supports up to 1M-token multimodal context, with strong reasoning, coding, and agentic tool use for large-scale deployment.

  🖼️Image generation💻Coding📷Images💬Chatting
  NewMultimodal

  Released 1mo ago
• Gen 3LongCat
  LongCat Flash Omni

  By Meituan

  560B parameter omni-modal MoE model (about 27B active) for real time audio-visual interaction, built on LongCat-Flash with multimodal perception and speech modules.

  🏋️Fitness📷Images💬Chatting🎥Videos
  NewMultimodal

  Released 1mo ago
• Gen 7Qwen
  Qwen3 Max Instruct

  By Alibaba

  Qwen3 Max Instruct is the trillion-parameter chat and tool-use version of Qwen3-Max, ranking near the top of LMArena and scoring about 69.6 on SWE-Bench Verified and 74.8 on Tau2 Bench for real-world coding and agent tasks.

  💻Coding💬Chatting🤖Agents🚀Productivity
  NewText

  Released 2mo ago
• Gen 3
  PersonaPlex 7B v1

  By NVIDIA

  PersonaPlex is NVIDIA’s full duplex conversational AI that lets you pick any role and voice, then maintains that persona through natural, low latency dialogue with interruptions, backchannels and realistic conversational rhythm.

  💬Chatting
  NewMultimodal

  Released 2mo ago
• Gen 7LFM
  LFM2.5 1.2B

  By Liquid AI

  LFM2.5-1.2B is Liquid AI’s 1.2B hybrid edge model for on-device chat and tools, trained on 28T tokens and RL, running under 1 GB of RAM while rivaling much larger LLMs on reasoning and agentic tasks.

  💬Chatting💻Coding🤖Chatbots🚀Productivity
  NewText

  Released 2mo ago
• Gen 7
  Nemotron 3 Nano 30b

  By NVIDIA

  Nemotron-3-Nano-30B-A3B is NVIDIA’s 30B-class LLM trained from scratch as a unified model for both reasoning and standard chat tasks. It can be configured to produce a reasoning trace before the final answer, letting developers switch reasoning behavior through the chat template.

  💬Chatting💻Coding🔍Advanced reasoning💻Conversational coding
  Text

  Released 3mo ago
• Gen 3Gemini
  Gemini 3 Flash

  By Google

  Gemini 3 Flash is Google’s fast, cost-efficient Gemini 3 model that pairs frontier-level multimodal reasoning with Flash latency, powering coding, analysis and chat and serving as default in the Gemini app and AI Mode in Search.

  💻Coding📊Data analysis📷Images💬Chatting
  Multimodal

  Released 3mo ago
• Gen 3GPT
  GPT 5.2 Instant

  By OpenAI

  GPT-5.2 Instant is the fast, lower-latency member of the 5.2 family, built for everyday Q&A, how-tos, research and writing where quick answers matter more than maximum depth of reasoning.

  📝Writing🤖ChatGPT🤖Chatbots💬Chatting
  Multimodal

  Released 3mo ago
• Gen 3GPT
  GPT 5.2

  By OpenAI

  GPT-5.2 is OpenAI’s December 2025 frontier model series for professional and agentic work, improving general intelligence, coding, vision, tool use and long-context reasoning across the Instant, Thinking and Pro variants.

  📷Images💻Coding📝Writing💬Chatting
  Multimodal

  Released 3mo ago
• Gen 7Mistral
  Ministral 3 14B

  By Mistral AI

  The largest Ministral 3 model offers frontier text and vision capabilities comparable to larger 24B models. Edge-optimized for single GPU deployment (24GB VRAM in FP8), it delivers state-of-the-art performance for chat, document analysis, and complex reasoning tasks with multilingual support across 40+ languages.

  💬Chatting📋Document chat🎥Videos
  Text

  Released 3mo ago
• Gen 7Mistral
  Ministral 3 8B

  By Mistral AI

  Best-in-class text and vision model for edge deployment, optimized for single GPU operation with minimal footprint. Features interleaved sliding-window attention for efficient inference. Ideal for constrained environments, chat interfaces, image/document understanding, and balanced local deployment scenarios.

  📷Images📝Writing💬Chatting🎥Videos
  Text

  Released 3mo ago
• Gen 3
  Mistral 3 14b Instruct

  By Mistral AI

  Ministral 3 14B Instruct 2512 is Mistral’s largest Ministral 3 model, built as an efficient instruction model with vision capabilities. Mistral positions it as delivering frontier-level capability while staying compact enough for local or edge deployment.

  💬Chatting🌐Text translation🔍Image interpretation💻Conversational coding
  Multimodal

  Released 3mo ago
• Gen 4
  Dia 2

  By Nari Labs

  Dia2 is an open source streaming dialogue TTS model that generates speech in real time from partial text, supports audio conditioning for natural back and forth conversations, and ships 1B and 2B checkpoints under Apache 2.0.

  🔊Text to speech💬Chatting🎥Videos🗒Transcription
  Audio

  Released 4mo ago
• Gen 3Gemini
  Gemini 3

  By Google

  Gemini 3 is Google’s next-generation multimodal model for careful reasoning, coding, and long-context dialogue. It accepts text, images, and audio, supports tool and function calling, streams responses, and returns clean JSON for production workflows.

  💻Coding📷Images💬Chatting
  Multimodal

  Released 4mo ago
• Gen 7
  Hermes 4.3

  By Nous Research

  Hermes 4.3 is Nous Research’s 36B hybrid reasoning model, based on Seed-OSS-36B, offering long context (up to 512k) and very high helpfulness on RefusalBench while staying locally deployable.

  📝Writing💬Chatting🎥Videos💼Personal assistant
  Text

  Released 4mo ago
• Gen 7
  VibeThinker-1.5B

  By WeiboAI

  VibeThinker-1.5B is WeiboAI’s compact instruction model designed for fast, cost-efficient assistants and automation.

  📝Writing📚Summaries💬Chatting
  Text

  Released 4mo ago
• Gen 7
  Locai l1 large

  By Locai Labs

  Locai L1-Large is an open instruction-tuned model built on Qwen3-235B Instruct, trained with Locai’s Forget-Me-Not framework to boost alignment while retaining math and science strength. It reports state-of-the-art Arena Hard v2 results, safer behavior on AgentHarm, and ships under Apache-2.0 with easy Transformers and vLLM use.

  📝Writing💬Chatting🚀Productivity
  Text

  Released 4mo ago
• Gen 3
  Kimi K2 Thinking

  By Moonshot AI

  Kimi K2 Thinking is Moonshot AI’s open-source reasoning-oriented K2 variant, built for deep multi-step thinking and tool use. It uses native INT4 quantization, supports a 256k context window, and is aimed at long-horizon agentic workflows rather than fast reflex chat.

  💬Chatting💻Coding🔍Advanced reasoning🧩Complex problem solving
  Text

  Released 4mo ago
• Gen 3GPT
  GPT Realtime

  By OpenAI

  GPT-Realtime is a low-latency multimodal model for live assistants. It understands text, images, and audio, and replies with text or natural speech. It supports function calling, long context, and clean JSON, so you can build voice-first apps that feel instant.

  🔒Private conversations💬Chatting
  Multimodal

  Released 4mo ago
• Gen 7
  Granite 4.0 1B

  By IBM

  Granite-4.0-1B is IBM’s lightweight instruction model built for efficient agentic tasks like instruction following and tool calling.

  💬Chatting🔍Text extraction🗂️Multilingual summarization💻Conversational coding
  Text

  Released 4mo ago
• Gen 7
  nanochat

  By Andrej Karpathy

  Nanochat is Andrej Karpathy’s tiny end-to-end chat assistant project. It shows how to build a small Transformer chatbot from scratch with simple training, inference, and a minimal UI, meant for learning and quick experimentation.

  🤖Chatbots💬Chatting
  Text

  Released 5mo ago
• Gen 7Claude
  Claude Haiku 4.5

  By Anthropic

  Claude Haiku 4.5 is Anthropic’s fast, cost-efficient model. It handles everyday reasoning, coding, and multimodal inputs with long context, tool and function calling, and reliable JSON output, making it a strong default for high-throughput apps.

  💻Coding📞Customer support📝Writing💬Chatting
  Text

  Released 5mo ago
• Gen 3
  Llama 3.3 Nemotron Super 49B

  By NVIDIA

  Llama-3.3-Nemotron-Super-49B-v1.5 is NVIDIA’s upgraded 49B reasoning model derived from Meta Llama-3.3-70B-Instruct. It is post-trained for reasoning, chat preference alignment, and agentic tasks such as RAG and tool use.

  💬Chatting💻Coding🔍Advanced reasoning💻Conversational coding
  Text

  Released 5mo ago
• Gen 3Grok
  Grok 4 Fast

  By xAI

  Grok 4 Fast is xAI’s optimized variant of the Grok 4 family. It delivers the same reasoning backbone as Grok 4 Heavy but tuned for speed and efficiency, with lower latency and cost. It’s designed for high-throughput applications, real-time assistants, and workloads where responsiveness matters more than maximum depth.

  💬Chatting📞Customer support💻Coding🎥Videos
  Multimodal

  Released 6mo ago
• Gen 3Mistral
  Magistral Medium 1.2

  By Mistral AI

  Magistral Medium 1.2 is Mistral AI’s mid-tier reasoning model, designed to balance capability and efficiency. It delivers stronger analysis, coding, and multilingual performance than the Small variant while keeping inference practical, with support for long-context inputs, JSON outputs, and tool/function calling.

  💻Coding📊Data analysis📝Writing💬Chatting
  Text

  Released 6mo ago
• Gen 3Mistral
  Magistral Small 1.2

  By Mistral AI

  Magistral Small 1.2 is Mistral AI’s compact reasoning model, optimized for efficiency and cost-sensitive deployments. With strong instruction following, reliable reasoning, and structured JSON outputs, it’s suited for lightweight copilots, chat assistants, and automation workflows where speed and affordability are key.

  💬Chatting📞Customer support🤖Chatbots
  Text

  Released 6mo ago
• Gen 7LongCat
  LongCat Flash Chat

  By Meituan

  Chat-oriented LongCat-Flash variant, a 560B MoE language model with around 27B parameters active per token, tuned as a fast, non-thinking foundation for general and agentic tasks.

  💬Chatting🤖Chatbots🚀Productivity
  Text

  Released 6mo ago
• Gen 3
  Seed OSS 36B Instruct

  By ByteDance

  Seed-OSS-36B-Instruct is ByteDance Seed’s open-source 36B instruction model built for long-context work, reasoning, agents, and general-purpose developer use. ByteDance positions the Seed-OSS line as efficient for its scale, reporting strong open-benchmark results from 12T training tokens.

  💬Chatting💻Coding🔍Advanced reasoning🛠️Comprehensive assistance
  Multimodal

  Released 7mo ago
• Gen 7Nemotron
  Nemotron Nano 9B v2

  By NVIDIA

  NVIDIA Nemotron Nano 9B v2 is a compact general-purpose language model designed to handle both reasoning-heavy and standard instruction tasks, with optional “thinking budget” control for how much reasoning it performs.

  💬Chatting💻Coding📝Writing📝Content
  Text

  Released 7mo ago
• Gen 7Qwen
  Qwen3-30B-A3B

  By Alibaba

  Qwen3-30B-A3B is Alibaba’s open-weight MoE LLM (Apache-2.0) with 30.5B params and ~3.3B activated. It supports toggleable “thinking” vs. non-thinking modes, strong agent/tool calling, 100+ languages, and a 32K native context (≈131K with YaRN). New 2507 variants raise context to 256K.

  📝Writing🏋️Fitness💬Chatting🌪️Brainstorming
  Text

  Released 7mo ago
• Gen 7GLM
  GLM 4.5

  By Z.ai

  GLM 4.5 is Zhipu AI’s flagship general-purpose model, tuned for strong reasoning, coding, and bilingual Chinese-English work. It supports long context, tool/function calling, reliable JSON output, streaming, and VL variants for image understanding—ready for RAG, agents, and enterprise copilots.

  💻Coding📷Images💬Chatting📚Stories
  Text

  Released 7mo ago
• Gen 7Qwen
  Qwen3-235B-A22B

  By Alibaba

  Qwen3-235B-A22B is Alibaba’s flagship open-source MoE LLM (Apache-2.0): 235B total parameters with 22B activated (128 experts, 8 active). It uniquely toggles thinking vs. non-thinking modes, supports 100+ languages, and excels at agentic tool use. Context is 32K native (≈131K with YaRN); the Instruct-2507 variant offers 256K native and up to ~1M tokens.

  💬Chatting📝Writing🤖Agents
  Text

  Released 8mo ago
• Gen 3Grok
  Grok 4 Heavy

  By xAI

  Grok 4 Heavy is xAI’s most powerful Grok 4 variant—frontier-level reasoning with native tool use, real-time search, a 256K-token context window, and higher rate limits. It’s available via the SuperGrok Heavy tier and the xAI API.

  💻Coding📊Data analysis📝Writing💬Chatting
  Text

  Released 8mo ago
• Gen 7
  Kimi K2 Instruct

  By Moonshot AI

  Kimi-K2-Instruct-0905 is Moonshot AI’s upgraded Kimi K2 instruct model, a 1T-parameter MoE system with 32B active parameters. It is tuned for agentic coding, better real-world coding performance, and stronger frontend generation without the long-thinking behavior of the dedicated thinking variant

  💬Chatting🔍Code reviews🔍Advanced reasoning🖥️Frontend development💻Conversational coding
  Text

  Released 8mo ago
• Gen 7Mistral
  Magistral Small

  By Mistral AI

  Magistral Small is a compact model from Mistral’s specialized lineup, tuned for efficiency while still offering solid reasoning and coding capabilities. At a smaller footprint, it delivers quick instruction following, structured JSON outputs, and function/tool calling support—making it ideal for cost-sensitive, real-time copilots and automations.

  💻Coding🌆Backgrounds💬Chatting
  Text

  Released 9mo ago
• Gen 4
  OmniVinci

  By NVIDIA

  OmniVinci is NVIDIA’s 9B omni-modal LLM that jointly understands images, video, audio, and text, achieving strong cross-modal reasoning with only about 0.2T training tokens.

  🖼️Image generation📷Images💬Chatting🎥Videos
  Image

  Released 9mo ago