| #1 | Claude 5.1
Anthropic | Language Models | 98.0 | Frontier assistant model known for long-context reasoning, agentic coding, and careful instruction following. | Base |
| #2 | GPT-5.6
OpenAI | Language Models | 97.5 | General-purpose frontier model with strong tool use, multimodal input, and broad consumer reach. | Base |
| #3 | Gemini 3.8
Google DeepMind | Language Models | 96.9 | Natively multimodal model family spanning text, image, audio, and video with deep Google ecosystem integration. | Base |
| #4 | Kimi K3
Moonshot AI | Language Models | 96.3 | Long-context mixture-of-experts model with a strong reputation for agentic workflows and research tasks. | Base |
| #5 | Grok 4.6
xAI | Language Models | 95.8 | Real-time-aware model with strong reasoning benchmarks and tight integration with X. | Base |
| #6 | DeepSeek V4
DeepSeek | Language Models | 95.2 | Open-weight mixture-of-experts model offering frontier-class performance at low inference cost. | Base |
| #7 | Qwen 3.8
Alibaba Cloud | Language Models | 94.7 | Broad open-weight family with sizes from edge to data center and strong multilingual coverage. | Base |
| #8 | Llama 4
Meta | Language Models | 94.2 | Widely adopted open-weight model family with a large fine-tuning ecosystem. | Base |
| #9 | MiniMax M2
MiniMax | Language Models | 93.6 | Efficient long-context model optimized for agents and coding at a low price point. | Base |
| #10 | GLM-5.3
Zhipu AI | Language Models | 93.0 | Bilingual open-weight model with strong agentic and coding performance. | Base |
| #11 | GPT-6 Astra
OpenAI | Language Models | 92.5 | Next-generation research preview emphasizing autonomous multi-step task completion. | Base |
| #12 | Mistral L3
Mistral AI | Language Models | 92.0 | European frontier model with a focus on efficiency, enterprise deployment, and open weights. | Base |
| #13 | Command A
Cohere | Language Models | 91.4 | Enterprise-focused model built for retrieval-augmented generation and tool use. | Base |
| #14 | Phi-4
Microsoft | Language Models | 90.8 | Small language model that punches above its size on reasoning and math benchmarks. | Base |
| #15 | Sora 2
OpenAI | Video Generation | 90.3 | Text- and image-to-video model producing coherent, physically plausible clips with synchronized audio. | Sponsored |
| #16 | Nemotron 4
NVIDIA | Language Models | 89.8 | Open model family tuned for synthetic data generation and enterprise agents. | Base |
| #17 | Whisper v3
OpenAI | Speech & Audio | 89.2 | Robust open-source speech recognition and translation model supporting dozens of languages. | Base |
| #18 | FLUX.1
Black Forest Labs | Image Generation | 88.7 | High-fidelity text-to-image model with strong prompt adherence and typography. | Base |
| #19 | Gemma 3
Google DeepMind | Language Models | 88.1 | Lightweight open-weight models derived from Gemini research, suitable for on-device use. | Base |
| #20 | EXAONE 3.0
LG AI Research | Language Models | 87.5 | Bilingual Korean–English model with competitive instruction-following performance. | Base |
| #21 | Yi 1.5
01.AI | Language Models | 87.0 | Open-weight bilingual model family with strong coding and math variants. | Base |
| #22 | DBRX
Databricks | Language Models | 86.5 | Open mixture-of-experts model built for enterprise data platforms. | Base |
| #23 | Solar Pro 2
Upstage | Language Models | 85.9 | Compact single-GPU model designed for efficient enterprise deployment. | Base |
| #24 | Baichuan 4
Baichuan AI | Language Models | 85.3 | Chinese-language-first model family with strong domain performance in finance and healthcare. | Base |
| #25 | InternLM 3
Shanghai AI Lab | Language Models | 84.8 | Open research model with strong reasoning and long-context support. | Base |
| #26 | Step 2
StepFun | Language Models | 84.2 | Trillion-parameter-class mixture-of-experts model for general assistance. | Base |
| #27 | Doubao 1.5
ByteDance | Language Models | 83.7 | Consumer assistant model powering ByteDance products at very large scale. | Base |
| #28 | Seed 1.0
ByteDance Seed | Language Models | 83.2 | Research-oriented foundation model line with strong multimodal variants. | Base |
| #29 | Skywork AI
Kunlun Tech | Language Models | 82.6 | Open model family with notable reward-model and reasoning releases. | Base |
| #30 | Marco-o1
Alibaba | Language Models | 82.0 | Open reasoning model exploring chain-of-thought and search-based inference. | Base |
| #31 | Qwen-VL
Alibaba Cloud | Vision & Multimodal | 81.5 | Vision-language model with strong document, chart, and OCR understanding. | Base |
| #32 | Aquila 3
BAAI | Language Models | 81.0 | Open bilingual foundation model from the Beijing Academy of AI. | Sponsored |
| #33 | BlueLM 7B
vivo | Language Models | 80.4 | On-device-oriented model optimized for mobile assistants. | Base |
| #34 | Code Llama
Meta | Code Models | 79.8 | Open code model family supporting infilling and long-context code completion. | Base |
| #35 | StarCoder 2
BigCode | Code Models | 79.3 | Transparent, permissively licensed code model trained on The Stack v2. | Base |
| #36 | Magicoder
UIUC | Code Models | 78.8 | Instruction-tuned code model using open-source-seeded synthetic data. | Base |
| #37 | Replit Code
Replit | Code Models | 78.2 | Code completion model tuned for in-IDE latency and multi-language support. | Base |
| #38 | Stable Code
Stability AI | Code Models | 77.7 | Compact code model for completion and fill-in-the-middle tasks. | Base |
| #39 | DeepSeek Coder
DeepSeek | Code Models | 77.1 | Open code model family with strong repository-level completion. | Base |
| #40 | Llama Dream
Community | Language Models | 76.5 | Community fine-tune lineage of Llama focused on creative writing and roleplay. | Base |
| #41 | CogVideoX
Zhipu AI | Video Generation | 76.0 | Open text-to-video diffusion model with a growing research community. | Base |
| #42 | LLaVA-NeXT
LLaVA Team | Vision & Multimodal | 75.5 | Open vision-language model widely used as a research baseline. | Base |
| #43 | IDEFix
Hugging Face | Vision & Multimodal | 74.9 | Open multimodal model line for image-text understanding. | Base |
| #44 | CogVLM
Zhipu AI | Vision & Multimodal | 74.3 | Visual expert model with strong grounding and captioning. | Base |
| #45 | Pika 2.1
Pika | Video Generation | 73.8 | Creator-focused video generation with editing and scene-extension tools. | Base |
| #46 | Runway Gen-3
Runway | Video Generation | 73.2 | Professional video generation model used across film and advertising workflows. | Base |
| #47 | Kling 2.1
Kuaishou | Video Generation | 72.7 | High-motion video generation model with strong character consistency. | Base |
| #48 | Hunyuan
Tencent | Vision & Multimodal | 72.2 | Multimodal foundation family spanning text, image, video, and 3D. | Base |
| #49 | Vidu 3.0
Shengshu | Video Generation | 71.6 | Video generation model noted for multi-subject consistency. | Base |
| #50 | Emu 3.0
BAAI | Vision & Multimodal | 71.0 | Unified next-token-prediction model for images, text, and video. | Base |
| #51 | VideoLLaMA
DAMO Academy | Vision & Multimodal | 70.5 | Audio-visual language model for video understanding. | Base |
| #52 | Janus-Pro
DeepSeek | Vision & Multimodal | 70.0 | Unified understanding-and-generation multimodal model. | Base |
| #53 | Kosmos-2.5
Microsoft | Vision & Multimodal | 69.4 | Document-literate multimodal model for text-rich images. | Base |
| #54 | Kosmos-2
Microsoft | Vision & Multimodal | 68.8 | Grounded multimodal model linking text to image regions. | Sponsored |
| #55 | Florence-2
Microsoft | Computer Vision | 68.3 | Compact vision foundation model handling captioning, detection, and segmentation. | Base |
| #56 | Moondream 2
Moondream | Vision & Multimodal | 67.8 | Tiny vision-language model that runs on edge hardware. | Base |
| #57 | PaliGemma
Google DeepMind | Vision & Multimodal | 67.2 | Open vision-language model built on Gemma for fine-tuning. | Base |
| #58 | StyleTTS
Columbia University | Speech & Audio | 66.7 | Expressive open text-to-speech with style diffusion. | Base |
| #59 | Bark 2
Suno | Speech & Audio | 66.1 | Generative audio model producing speech, music, and sound effects. | Base |
| #60 | ElevenLabs
ElevenLabs | Speech & Audio | 65.5 | Production voice synthesis and cloning platform widely used in media. | Base |
| #61 | Fish Speech
Fish Audio | Speech & Audio | 65.0 | Open multilingual text-to-speech with few-shot voice cloning. | Base |
| #62 | GPT-4o Audio
OpenAI | Speech & Audio | 64.4 | Real-time speech-to-speech model powering conversational voice assistants. | Base |
| #63 | Moshi
Kyutai | Speech & Audio | 63.9 | Full-duplex real-time spoken dialogue model. | Base |
| #64 | VoiceCraft
UT Austin | Speech & Audio | 63.3 | Zero-shot speech editing and synthesis model. | Base |
| #65 | Fuyu
Adept | Vision & Multimodal | 62.8 | Simplified multimodal architecture for UI and chart understanding. | Base |
| #66 | DINOv2
Meta | Computer Vision | 62.2 | Self-supervised visual features widely used as a backbone. | Base |
| #67 | MobileVLM
Meituan | Vision & Multimodal | 61.7 | Vision-language model designed for mobile and embedded devices. | Sponsored |
| #68 | SAM 2
Meta | Computer Vision | 61.1 | Promptable segmentation model for images and video. | Base |
| #69 | CLIP
OpenAI | Computer Vision | 60.6 | Foundational image-text contrastive model used across the industry. | Base |
| #70 | EVA
BAAI | Computer Vision | 60.0 | Large-scale vision representation model with strong transfer performance. | Base |
| #71 | OpenCoder
INF / M-A-P | Code Models | 59.5 | Fully open code model with released data pipeline. | Base |
| #72 | WizardCoder
Microsoft | Code Models | 58.9 | Evol-Instruct-tuned code model with strong HumanEval results. | Base |
| #73 | CodeGeeX
Zhipu AI | Code Models | 58.4 | Multilingual code generation model with IDE plugins. | Base |
| #74 | InternVL
Shanghai AI Lab | Vision & Multimodal | 57.8 | Open vision-language model family competitive with proprietary systems. | Base |
| #75 | CogVideo
Tsinghua | Video Generation | 57.3 | Early large-scale open text-to-video transformer. | Base |
| #76 | VILA
NVIDIA | Vision & Multimodal | 56.8 | Visual language model optimized for efficient inference. | Base |
| #77 | NVLM
NVIDIA | Vision & Multimodal | 56.2 | Frontier-class open multimodal model from NVIDIA research. | Base |
| #78 | ShieldGemma
Google DeepMind | Language Models | 55.6 | Safety classifier models for moderating LLM inputs and outputs. | Base |
| #79 | Reka Core
Reka | Vision & Multimodal | 55.1 | Multimodal model with native video and audio understanding. | Base |
| #80 | Mamba
Together / CMU | Language Models | 54.5 | State-space sequence model offering linear-time inference. | Base |
| #81 | Titan
Amazon | Language Models | 54.0 | Amazon's foundation model family available through Bedrock. | Base |
| #82 | Mixbread
Mixedbread | Embeddings & Retrieval | 53.4 | Open embedding and reranking models for retrieval. | Base |
| #83 | Cohere Embed
Cohere | Embeddings & Retrieval | 52.9 | Multilingual embedding model built for enterprise search. | Base |
| #84 | BGE M3
BAAI | Embeddings & Retrieval | 52.3 | Multi-functional, multilingual, multi-granularity embedding model. | Base |
| #85 | Voyage 3
Voyage AI | Embeddings & Retrieval | 51.8 | Domain-tuned embedding models for code, legal, and finance. | Base |
| #86 | ColBERTv2
Stanford | Embeddings & Retrieval | 51.2 | Late-interaction retrieval model with strong precision. | Base |
| #87 | Contriever
Meta | Embeddings & Retrieval | 50.7 | Unsupervised dense retriever used as a research baseline. | Base |
| #88 | Jina Embed
Jina AI | Embeddings & Retrieval | 50.1 | Long-context embedding models with multilingual support. | Base |
| #89 | GTE Large
Alibaba DAMO | Embeddings & Retrieval | 49.6 | General text embeddings with strong MTEB performance. | Base |
| #90 | BAAI bge
BAAI | Embeddings & Retrieval | 49.0 | Widely adopted open embedding family for semantic search. | Base |
| #91 | InstructorXL
HKU / UW | Embeddings & Retrieval | 48.5 | Instruction-conditioned text embeddings. | Base |
| #92 | Snowflake Arctic
Snowflake | Language Models | 47.9 | Enterprise-focused open model and embedding family. | Base |
| #93 | Tortoise
neonbjb | Speech & Audio | 47.4 | High-quality open text-to-speech favoring realism over speed. | Base |
| #94 | XTTS
Coqui | Speech & Audio | 46.8 | Multilingual voice-cloning text-to-speech model. | Base |
| #95 | MPT-7B
MosaicML | Language Models | 46.3 | Commercially usable open model with long-context variants. | Base |
| #96 | Falcon 3
TII | Language Models | 45.7 | Open model family from the UAE with efficient small variants. | Base |
| #97 | Phi-3 Mini
Microsoft | Language Models | 45.2 | 3.8B-parameter model that runs well on phones and laptops. | Base |
| #98 | Stable Audio
Stability AI | Speech & Audio | 44.6 | Text-to-audio model for music and sound effect generation. | Base |
| #99 | Luminous
Aleph Alpha | Language Models | 44.1 | European sovereign-AI model family with explainability features. | Base |
| #100 | OLMo 2
Allen Institute for AI | Language Models | 43.5 | Fully open model with released training data, code, and checkpoints. | Base |