A curated snapshot of high-impact AI and LLM models, grouped by practical use case.

1. General-Purpose LLMs (Text + Reasoning)

The all-round models for chat, reasoning, analysis, and agents.

Model Provider Type Notes
GPT-5 / GPT-5 Pro / GPT-5 mini OpenAI Closed State-of-the-art general capability plus strong reasoning.
o3 / o4-mini OpenAI Closed Deep reasoning models.
Claude 4.5 Sonnet / Claude 4.1 Opus / Haiku 4.5 Anthropic Closed Leader in coding, agents, and long-context tasks.
Gemini 2.5 Pro / Flash / Flash-Lite Google DeepMind Closed Native multimodal with 1M+ token context.
Grok 4 / Grok 4 Heavy xAI Closed Reasoning plus real-time X access.
Llama 4 (Scout, Maverick, Behemoth) Meta Open Native multimodal MoE family.
DeepSeek V3.2 / DeepSeek-R1 DeepSeek Open Efficient MoE; R1 is reasoning-focused.
Qwen3 / Qwen3-Max Alibaba Open (mostly) Very strong multilingual and agent performance.
Mistral Large 2 / Medium 3 Mistral AI Mixed Strong function calling with European ecosystem focus.
Command A / R+ Cohere Closed Enterprise-focused, strong for RAG and multilingual use.
Nova Pro / Premier Amazon Closed Deep integration with Bedrock.
Phi-4 Microsoft Open Very capable SLM for its size.
Gemma 3 Google Open Google’s open model family.

2. Deep Reasoning Models (Reasoning / Thinking)

Model Provider Type Notes
OpenAI o3 / o4-mini OpenAI Closed Long internal reasoning traces.
Claude 4.1 Opus (extended thinking) Anthropic Closed Deep-think mode.
Gemini 2.5 Pro Deep Think Google Closed Parallelized reasoning.
DeepSeek-R1 / R1-0528 DeepSeek Open Open-source reference in reasoning.
Qwen3-Thinking / QwQ-32B Alibaba Open Open-source reasoning models.
Grok 4 Heavy xAI Closed Multi-agent reasoning.
Kimi K2 Moonshot AI Open Reasoning plus agent workflows, open-source.
GLM-4.6 Zhipu AI Open Open-source with strong reasoning quality.

3. Code-Specialized Models

Model Provider Type Notes
Claude 4.5 Sonnet Anthropic Closed De facto leader in coding and code-agent benchmarks.
GPT-5 Codex OpenAI Closed Optimized for software engineering workflows.
Gemini 2.5 Pro Google Closed Very strong long-context coding.
DeepSeek-Coder V3 / V2 DeepSeek Open Open-source reference for code models.
Qwen3-Coder Alibaba Open Top open-source coding model line.
Codestral 25.08 Mistral Closed Purpose-built for coding.
Code Llama / Llama 4 Code Meta Open Open models for coding tasks.
StarCoder 2 BigCode / HuggingFace Open Open code model family.
GitHub Copilot models GitHub/OpenAI Closed Deep IDE integration.
Cursor Composer / Tab models Cursor Closed In-house tuned models for editor-native workflows.

4. Image Generation

Model Provider Type Notes
GPT-Image-1 / DALL-E 3 OpenAI Closed Integrated in ChatGPT workflows.
Imagen 4 / Imagen 4 Ultra Google Closed High photorealism and quality.
Nano Banana (Gemini 2.5 Flash Image) Google Closed Viral conversational image editing.
Midjourney v7 Midjourney Closed Leading artistic aesthetics.
Firefly Image 4 Adobe Closed Trained with clean licensed data.
Ideogram 3.0 Ideogram Closed Excellent text rendering in images.
FLUX.1.1 Pro / FLUX.2 Black Forest Labs Mixed FLUX.1 [dev] is open.
Stable Diffusion 3.5 / SDXL Stability AI Open Open-source reference stack.
Recraft V3 Recraft Closed Strong for vector and graphic design.
HiDream / Qwen-Image Alibaba Open Fast-evolving open Chinese image models.

5. Video Generation

Model Provider Type Notes
Sora 2 / Sora 2 Pro OpenAI Closed Video generation with synchronized audio.
Veo 3 / Veo 3.1 Google DeepMind Closed Native video + audio, top visual quality.
Runway Gen-4 / Gen-4 Turbo Runway Closed Professional film-making workflows.
Kling 2.1 / 2.5 Kuaishou Closed Highly realistic Chinese leader.
Hailuo 02 MiniMax Closed Cinematic video generation.
Pika 2.2 Pika Labs Closed Creative effects and stylized generation.
Luma Ray 2 / Dream Machine Luma AI Closed Physically consistent scene behavior.
Wan 2.2 / Wan 2.5 Alibaba Open Top open-source video line.
HunyuanVideo Tencent Open High-quality open-source video model.
Mochi 1 Genmo Open Early open-source video pioneer.
LTX Video Lightricks Open Fast and efficient generation.
Marey Moonvalley Closed Trained on licensed datasets.

6. Audio, Voice, and Music

Voice (TTS / STT / Conversational)

Model Provider Type Notes
GPT-4o Realtime / GPT Realtime OpenAI Closed Low-latency conversational voice.
Gemini Live Google Closed Real-time multimodal voice.
ElevenLabs v3 / Eleven Multilingual v2 ElevenLabs Closed TTS category leader.
Cartesia Sonic 2 Cartesia Closed Ultra-fast TTS.
OpenAI Whisper v3 OpenAI Open Open-source STT standard.
NVIDIA Canary / Parakeet NVIDIA Open Open-source STT models.
Sesame CSM Sesame Open Open-source natural voice generation.
Kyutai Moshi / Unmute Kyutai Open Open-source full-duplex voice models.

Music

Model Provider Type Notes
Suno v4.5 / v5 Suno Closed Closed, category-leading music generation.
Udio v1.5 Udio Closed Closed model family.
Lyria 2 Google DeepMind Closed Closed music generation line.
MusicGen / AudioCraft Meta Open Open-source music generation stack.
Stable Audio 2.5 Stability AI Mixed Mixed licensing model.
ACE-Step StepFun Open Open-source.

7. Multimodal / Vision Language Models (VLM)

Model Provider Type Notes
GPT-5 / GPT-4o OpenAI Closed Vision + text + audio stack.
Gemini 2.5 Pro Google Closed Native multimodal support across image, video, and audio.
Claude 4.5 Sonnet (vision) Anthropic Closed Vision plus computer-use workflows.
Llama 4 (multimodal) Meta Open Open multimodal family.
Qwen3-VL / Qwen2.5-VL Alibaba Open Open VLM reference models.
InternVL 3 Shanghai AI Lab Open Open multimodal model.
Pixtral Large Mistral Mixed Mixed approach with strong vision capabilities.
Molmo Allen AI Open Open multimodal line.
DeepSeek-VL2 DeepSeek Open Open VLM family.

8. Agent / Computer-Use / Web Action Models

Model Provider Type Notes
Claude 4.5 Sonnet (computer use) Anthropic Closed Controls screen, mouse, and keyboard.
GPT-5 + Operator OpenAI Closed Web-navigating autonomous agent workflows.
Gemini 2.5 Computer Use Google Closed Google equivalent for interactive computer-use agents.
Manus Butterfly Effect Closed General autonomous agent platform.
Magma Microsoft Open Open-source multimodal agent model.
UI-TARS ByteDance Open Open-source GUI control model.
OpenAI Agents SDK + Responses API OpenAI Closed Framework plus model runtime for agent systems.

9. Scientific / Domain-Specific Models

Domain Model Type Notes
Biology / Proteins AlphaFold 3 (Google DeepMind), ESM3 (EvolutionaryScale), Boltz-2 (MIT, open-source), Chai-1 Mixed Protein folding and design.
Chemistry / Materials GNoME (Google), MatterGen (Microsoft) Closed Material discovery and generation.
Medicine Med-Gemini, Med-PaLM 2 (Google), MedLM Closed Clinical and healthcare-oriented models.
Genomics Evo 2 (Arc Institute, open-source) Open DNA and genomic foundation model.
Climate / Weather GraphCast, GenCast (Google), Aurora (Microsoft), Pangu-Weather (Huawei) Mixed Weather and climate prediction.
Robotics Gemini Robotics 1.5, pi0 (Physical Intelligence), Helix (Figure), GR00T N1 (NVIDIA, open-source), RT-2 (Google) Mixed Vision-Language-Action model families.
Mathematics AlphaProof, AlphaGeometry 2 (Google), DeepSeek-Prover V2 (open-source) Mixed Formal reasoning and theorem proving.
Finance BloombergGPT, FinGPT (open-source) Mixed Finance-focused language models.
Legal Harvey (GPT-based), Paxton AI Closed Legal assistant systems.

10. 3D / World Models / Simulation

Model Provider Type Notes
Genie 3 Google DeepMind Closed Interactive world model.
Cosmos NVIDIA Open Open world foundation models.
Hunyuan3D 2.1 Tencent Open Open 3D generation model.
TRELLIS Microsoft Open Open image-to-3D pipeline.
Meshy 5 / Rodin Meshy / Deemos Closed Closed commercial 3D models.
V-JEPA 2 Meta Open Open world model line.

11. Embeddings and Reranking (Core for RAG / Agents)

Model Provider Type Notes
text-embedding-3-large OpenAI Closed Commercial standard embedding model.
Voyage 3 / voyage-3-large Voyage AI (Anthropic) Closed Top embedding quality in practice.
Cohere Embed v4 Cohere Closed Multilingual and multimodal embedding stack.
Gemini Embedding Google Closed Native integration in Vertex AI.
BGE-M3 / BGE-Gemma2 BAAI Open Open-source embedding references.
Nomic Embed v2 Nomic Open Open embedding model family.
Jina Embeddings v3 Jina AI Open Open high-quality embeddings.
Qwen3 Embedding / Reranker Alibaba Open Open embedding and reranker models.

12. Small Language Models (SLMs) for Edge / On-Device

Model Provider Type Notes
Phi-4 / Phi-4-mini Microsoft Open Open SLM family.
Gemma 3 (1B-27B) Google Open Open and scalable SLM line.
Llama 3.2 (1B/3B) Meta Open Open edge-friendly models.
Qwen3 (0.6B-4B) Alibaba Open Open compact model range.
SmolLM 3 HuggingFace Open Open lightweight language model.
Apple Intelligence Foundation Models Apple Closed On-device iOS/macOS inference.
Gemini Nano Google Closed On-device Android/Chrome stack.
Ministral 3B / 8B Mistral Closed Edge-focused deployment options.