| { |
| "dataset": "Convly AI Models Database", |
| "description": "Hand-verified specifications, licenses, API pricing (USD per 1M tokens) and local-hardware VRAM requirements for frontier and open-weight AI models.", |
| "version": "2026-07-03", |
| "source": "https://convly.ai/models/", |
| "license": "CC BY 4.0", |
| "attribution": "Convly.ai (https://convly.ai) — link to https://convly.ai/models/", |
| "model_count": 30, |
| "fields": [ |
| "name", |
| "slug", |
| "convly_url", |
| "developer", |
| "model_type", |
| "modality", |
| "parameters", |
| "context_window", |
| "max_output", |
| "license", |
| "open_weights", |
| "release_date", |
| "input_price", |
| "output_price", |
| "api_providers", |
| "vram_q4", |
| "min_gpu", |
| "official_url" |
| ], |
| "models": [ |
| { |
| "name": "Claude Fable 5", |
| "slug": "claude-fable-5", |
| "convly_url": "https://convly.ai/model/claude-fable-5/", |
| "developer": "Anthropic", |
| "model_type": "LLM (frontier reasoning)", |
| "modality": "Text, Vision → Text", |
| "parameters": "Undisclosed", |
| "context_window": "1M", |
| "max_output": "128K", |
| "license": "Proprietary", |
| "open_weights": "no", |
| "release_date": "2026", |
| "input_price": 10, |
| "output_price": 50, |
| "api_providers": "Anthropic, AWS", |
| "vram_q4": null, |
| "min_gpu": null, |
| "official_url": "https://www.anthropic.com/claude" |
| }, |
| { |
| "name": "Claude Haiku 4.5", |
| "slug": "claude-haiku-4-5", |
| "convly_url": "https://convly.ai/model/claude-haiku-4-5/", |
| "developer": "Anthropic", |
| "model_type": "LLM", |
| "modality": "Text, Vision → Text", |
| "parameters": "Undisclosed", |
| "context_window": "200K", |
| "max_output": "64K", |
| "license": "Proprietary", |
| "open_weights": "no", |
| "release_date": "2025", |
| "input_price": 1, |
| "output_price": 5, |
| "api_providers": "Anthropic, AWS, Vertex AI, Azure", |
| "vram_q4": null, |
| "min_gpu": null, |
| "official_url": "https://www.anthropic.com/claude" |
| }, |
| { |
| "name": "Claude Opus 4.8", |
| "slug": "claude-opus-4-8", |
| "convly_url": "https://convly.ai/model/claude-opus-4-8/", |
| "developer": "Anthropic", |
| "model_type": "LLM (reasoning)", |
| "modality": "Text, Vision → Text", |
| "parameters": "Undisclosed", |
| "context_window": "1M", |
| "max_output": "128K", |
| "license": "Proprietary", |
| "open_weights": "no", |
| "release_date": "2026", |
| "input_price": 5, |
| "output_price": 25, |
| "api_providers": "Anthropic, AWS, Vertex AI, Azure", |
| "vram_q4": null, |
| "min_gpu": null, |
| "official_url": "https://www.anthropic.com/claude" |
| }, |
| { |
| "name": "Claude Sonnet 4.6", |
| "slug": "claude-sonnet-4-6", |
| "convly_url": "https://convly.ai/model/claude-sonnet-4-6/", |
| "developer": "Anthropic", |
| "model_type": "LLM", |
| "modality": "Text, Vision → Text", |
| "parameters": "Undisclosed", |
| "context_window": "1M", |
| "max_output": "64K", |
| "license": "Proprietary", |
| "open_weights": "no", |
| "release_date": "2025", |
| "input_price": 3, |
| "output_price": 15, |
| "api_providers": "Anthropic, AWS, Vertex AI, Azure", |
| "vram_q4": null, |
| "min_gpu": null, |
| "official_url": "https://www.anthropic.com/claude" |
| }, |
| { |
| "name": "DeepSeek R1", |
| "slug": "deepseek-r1", |
| "convly_url": "https://convly.ai/model/deepseek-r1/", |
| "developer": "DeepSeek", |
| "model_type": "LLM (MoE, reasoning)", |
| "modality": "Text → Text", |
| "parameters": "671B total / 37B active (MoE)", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "MIT (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.5, |
| "output_price": 2.15, |
| "api_providers": "DeepSeek, DeepInfra, OpenRouter", |
| "vram_q4": "~400 GB", |
| "min_gpu": "Multi-GPU server", |
| "official_url": "https://huggingface.co/deepseek-ai" |
| }, |
| { |
| "name": "DeepSeek R1 Distill Llama 70B", |
| "slug": "deepseek-r1-distill-llama-70b", |
| "convly_url": "https://convly.ai/model/deepseek-r1-distill-llama-70b/", |
| "developer": "DeepSeek", |
| "model_type": "LLM (dense, reasoning)", |
| "modality": "Text → Text", |
| "parameters": "70B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "MIT (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.8, |
| "output_price": 0.8, |
| "api_providers": "DeepInfra, OpenRouter, Ollama", |
| "vram_q4": "~40 GB", |
| "min_gpu": "2× RTX 4090 / 1× 48GB", |
| "official_url": "https://huggingface.co/deepseek-ai" |
| }, |
| { |
| "name": "DeepSeek V4-Flash", |
| "slug": "deepseek-v4-flash", |
| "convly_url": "https://convly.ai/model/deepseek-v4-flash/", |
| "developer": "DeepSeek", |
| "model_type": "LLM (MoE)", |
| "modality": "Text → Text", |
| "parameters": "284B total / ~13B active (MoE)", |
| "context_window": "1M", |
| "max_output": "384K", |
| "license": "MIT (open)", |
| "open_weights": "yes", |
| "release_date": "2026-04", |
| "input_price": 0.14, |
| "output_price": 0.28, |
| "api_providers": "DeepSeek, OpenRouter", |
| "vram_q4": "~140 GB", |
| "min_gpu": "2× H100 80GB (4-bit)", |
| "official_url": "https://huggingface.co/deepseek-ai" |
| }, |
| { |
| "name": "DeepSeek V4-Pro", |
| "slug": "deepseek-v4-pro", |
| "convly_url": "https://convly.ai/model/deepseek-v4-pro/", |
| "developer": "DeepSeek", |
| "model_type": "LLM (MoE)", |
| "modality": "Text → Text", |
| "parameters": "1.6T total / ~49B active (MoE)", |
| "context_window": "1M", |
| "max_output": "384K", |
| "license": "MIT (open)", |
| "open_weights": "yes", |
| "release_date": "2026-04", |
| "input_price": 0.435, |
| "output_price": 0.87, |
| "api_providers": "DeepSeek, OpenRouter", |
| "vram_q4": "~800 GB", |
| "min_gpu": "Multi-GPU server (e.g. 8× H100 80GB)", |
| "official_url": "https://huggingface.co/deepseek-ai" |
| }, |
| { |
| "name": "Gemini 3.1 Pro", |
| "slug": "gemini-3-1-pro", |
| "convly_url": "https://convly.ai/model/gemini-3-1-pro/", |
| "developer": "Google", |
| "model_type": "LLM (multimodal)", |
| "modality": "Text, Image, Audio, Video → Text", |
| "parameters": "Undisclosed", |
| "context_window": "1.05M", |
| "max_output": "65K", |
| "license": "Proprietary", |
| "open_weights": "no", |
| "release_date": "2026", |
| "input_price": 2, |
| "output_price": 12, |
| "api_providers": "Google AI Studio, Vertex AI", |
| "vram_q4": null, |
| "min_gpu": null, |
| "official_url": "https://ai.google.dev/" |
| }, |
| { |
| "name": "Gemini 3.5 Flash", |
| "slug": "gemini-3-5-flash", |
| "convly_url": "https://convly.ai/model/gemini-3-5-flash/", |
| "developer": "Google", |
| "model_type": "LLM (multimodal)", |
| "modality": "Text, Image, Audio, Video → Text", |
| "parameters": "Undisclosed", |
| "context_window": "1M", |
| "max_output": "65K", |
| "license": "Proprietary", |
| "open_weights": "no", |
| "release_date": "2026", |
| "input_price": 1.5, |
| "output_price": 9, |
| "api_providers": "Google AI Studio, Vertex AI", |
| "vram_q4": null, |
| "min_gpu": null, |
| "official_url": "https://ai.google.dev/" |
| }, |
| { |
| "name": "Gemma 3 12B", |
| "slug": "gemma-3-12b", |
| "convly_url": "https://convly.ai/model/gemma-3-12b/", |
| "developer": "Google", |
| "model_type": "LLM (multimodal)", |
| "modality": "Text, Image → Text", |
| "parameters": "12B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Gemma (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.05, |
| "output_price": 0.15, |
| "api_providers": "Google AI Studio, OpenRouter, Ollama", |
| "vram_q4": "~8 GB", |
| "min_gpu": "RTX 4070 12GB", |
| "official_url": "https://huggingface.co/google" |
| }, |
| { |
| "name": "Gemma 3 27B", |
| "slug": "gemma-3-27b", |
| "convly_url": "https://convly.ai/model/gemma-3-27b/", |
| "developer": "Google", |
| "model_type": "LLM (multimodal)", |
| "modality": "Text, Image → Text", |
| "parameters": "27B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Gemma (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.08, |
| "output_price": 0.16, |
| "api_providers": "Google AI Studio, OpenRouter, Ollama", |
| "vram_q4": "~16 GB", |
| "min_gpu": "RTX 4080 16GB / RTX 4090", |
| "official_url": "https://huggingface.co/google" |
| }, |
| { |
| "name": "Gemma 3 4B", |
| "slug": "gemma-3-4b", |
| "convly_url": "https://convly.ai/model/gemma-3-4b/", |
| "developer": "Google", |
| "model_type": "LLM (multimodal)", |
| "modality": "Text, Image → Text", |
| "parameters": "4B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Gemma (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.05, |
| "output_price": 0.1, |
| "api_providers": "Google AI Studio, Ollama", |
| "vram_q4": "~3 GB", |
| "min_gpu": "Any 6GB+ GPU", |
| "official_url": "https://huggingface.co/google" |
| }, |
| { |
| "name": "GLM 5.2", |
| "slug": "glm-5-2", |
| "convly_url": "https://convly.ai/model/glm-5-2/", |
| "developer": "Zhipu AI", |
| "model_type": "LLM (coding/agentic, MoE)", |
| "modality": "Text → Text", |
| "parameters": "744B total / ~40B active (MoE)", |
| "context_window": "1M", |
| "max_output": "131K", |
| "license": "MIT (open)", |
| "open_weights": "yes", |
| "release_date": "2026-06", |
| "input_price": 1.4, |
| "output_price": 4.4, |
| "api_providers": "Zhipu (Z.ai), OpenRouter", |
| "vram_q4": "~370 GB", |
| "min_gpu": "Multi-GPU server (e.g. 5× H100 80GB)", |
| "official_url": "https://github.com/zai-org/GLM-5" |
| }, |
| { |
| "name": "GPT-5.5", |
| "slug": "gpt-5-5", |
| "convly_url": "https://convly.ai/model/gpt-5-5/", |
| "developer": "OpenAI", |
| "model_type": "LLM (reasoning)", |
| "modality": "Text, Vision → Text", |
| "parameters": "Undisclosed", |
| "context_window": "1.05M", |
| "max_output": "128K", |
| "license": "Proprietary", |
| "open_weights": "no", |
| "release_date": "2026", |
| "input_price": 5, |
| "output_price": 30, |
| "api_providers": "OpenAI, Azure", |
| "vram_q4": null, |
| "min_gpu": null, |
| "official_url": "https://openai.com/api/" |
| }, |
| { |
| "name": "Kimi K2.7 Code", |
| "slug": "kimi-k2-7-code", |
| "convly_url": "https://convly.ai/model/kimi-k2-7-code/", |
| "developer": "Moonshot AI", |
| "model_type": "LLM (coding, MoE)", |
| "modality": "Text → Text", |
| "parameters": "1T total / 32B active (MoE)", |
| "context_window": "256K", |
| "max_output": "—", |
| "license": "Modified MIT (open)", |
| "open_weights": "yes", |
| "release_date": "2026-06", |
| "input_price": 0.6, |
| "output_price": 2.5, |
| "api_providers": "Moonshot, OpenRouter", |
| "vram_q4": "~500 GB", |
| "min_gpu": "Multi-GPU server", |
| "official_url": "https://huggingface.co/moonshotai" |
| }, |
| { |
| "name": "Llama 3.1 8B", |
| "slug": "llama-3-1-8b", |
| "convly_url": "https://convly.ai/model/llama-3-1-8b/", |
| "developer": "Meta", |
| "model_type": "LLM (dense)", |
| "modality": "Text → Text", |
| "parameters": "8B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Llama 3.1 Community (open)", |
| "open_weights": "yes", |
| "release_date": "2024", |
| "input_price": 0.02, |
| "output_price": 0.03, |
| "api_providers": "Together, DeepInfra, OpenRouter, Ollama", |
| "vram_q4": "~5 GB", |
| "min_gpu": "Any 8GB GPU", |
| "official_url": "https://www.llama.com/" |
| }, |
| { |
| "name": "Llama 3.3 70B", |
| "slug": "llama-3-3-70b", |
| "convly_url": "https://convly.ai/model/llama-3-3-70b/", |
| "developer": "Meta", |
| "model_type": "LLM (dense)", |
| "modality": "Text → Text", |
| "parameters": "70B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Llama 3.3 Community (open)", |
| "open_weights": "yes", |
| "release_date": "2024", |
| "input_price": 0.1, |
| "output_price": 0.32, |
| "api_providers": "Together, DeepInfra, OpenRouter, Ollama", |
| "vram_q4": "~40 GB", |
| "min_gpu": "2× RTX 4090 / 1× 48GB", |
| "official_url": "https://www.llama.com/" |
| }, |
| { |
| "name": "Llama 4 Maverick", |
| "slug": "llama-4-maverick", |
| "convly_url": "https://convly.ai/model/llama-4-maverick/", |
| "developer": "Meta", |
| "model_type": "Multimodal (MoE)", |
| "modality": "Text, Image → Text", |
| "parameters": "400B total / 17B active (MoE)", |
| "context_window": "1M", |
| "max_output": null, |
| "license": "Llama 4 Community (EU-restricted)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.15, |
| "output_price": 0.6, |
| "api_providers": "Meta, Together, OpenRouter", |
| "vram_q4": "~240 GB", |
| "min_gpu": "Multi-GPU server", |
| "official_url": "https://www.llama.com/models/llama-4/" |
| }, |
| { |
| "name": "Llama 4 Scout", |
| "slug": "llama-4-scout", |
| "convly_url": "https://convly.ai/model/llama-4-scout/", |
| "developer": "Meta", |
| "model_type": "Multimodal (MoE)", |
| "modality": "Text, Image → Text", |
| "parameters": "109B total / 17B active (MoE)", |
| "context_window": "10M", |
| "max_output": null, |
| "license": "Llama 4 Community (EU-restricted)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.1, |
| "output_price": 0.3, |
| "api_providers": "Meta, Together, OpenRouter", |
| "vram_q4": "~65 GB", |
| "min_gpu": "H100 80GB / Mac 128GB", |
| "official_url": "https://www.llama.com/models/llama-4/" |
| }, |
| { |
| "name": "Mistral 7B", |
| "slug": "mistral-7b", |
| "convly_url": "https://convly.ai/model/mistral-7b/", |
| "developer": "Mistral AI", |
| "model_type": "LLM (dense)", |
| "modality": "Text → Text", |
| "parameters": "7B", |
| "context_window": "32K", |
| "max_output": null, |
| "license": "Apache 2.0 (open)", |
| "open_weights": "yes", |
| "release_date": "2023", |
| "input_price": 0.02, |
| "output_price": 0.03, |
| "api_providers": "Mistral, DeepInfra, OpenRouter, Ollama", |
| "vram_q4": "~4.5 GB", |
| "min_gpu": "Any 6GB GPU", |
| "official_url": "https://huggingface.co/mistralai" |
| }, |
| { |
| "name": "Mistral Large 3", |
| "slug": "mistral-large-3", |
| "convly_url": "https://convly.ai/model/mistral-large-3/", |
| "developer": "Mistral AI", |
| "model_type": "LLM (MoE)", |
| "modality": "Text → Text", |
| "parameters": "675B total / 41B active (MoE)", |
| "context_window": "256K", |
| "max_output": null, |
| "license": "Apache 2.0 (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 2, |
| "output_price": 6, |
| "api_providers": "Mistral, OpenRouter", |
| "vram_q4": "~400 GB", |
| "min_gpu": "Multi-GPU server", |
| "official_url": "https://huggingface.co/mistralai" |
| }, |
| { |
| "name": "Mistral NeMo 12B", |
| "slug": "mistral-nemo-12b", |
| "convly_url": "https://convly.ai/model/mistral-nemo-12b/", |
| "developer": "Mistral AI", |
| "model_type": "LLM (dense)", |
| "modality": "Text → Text", |
| "parameters": "12B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Apache 2.0 (open)", |
| "open_weights": "yes", |
| "release_date": "2024", |
| "input_price": 0.02, |
| "output_price": 0.04, |
| "api_providers": "Mistral, OpenRouter, Ollama", |
| "vram_q4": "~7.5 GB", |
| "min_gpu": "RTX 4070 12GB / RTX 3060", |
| "official_url": "https://huggingface.co/mistralai" |
| }, |
| { |
| "name": "NVIDIA Nemotron 3 Nano Omni", |
| "slug": "nvidia-nemotron-3-nano-omni", |
| "convly_url": "https://convly.ai/model/nvidia-nemotron-3-nano-omni/", |
| "developer": "NVIDIA", |
| "model_type": "Multimodal (omni)", |
| "modality": "Text, Image, Audio, Video → Text", |
| "parameters": "30B total / ~3B active (MoE)", |
| "context_window": "256K", |
| "max_output": "—", |
| "license": "NVIDIA Open Model Agreement", |
| "open_weights": "yes", |
| "release_date": "2026", |
| "input_price": null, |
| "output_price": null, |
| "api_providers": "Hugging Face, OpenRouter, NVIDIA NIM", |
| "vram_q4": "~21 GB (NVFP4)", |
| "min_gpu": "RTX 5090 32GB (NVFP4) / H100 80GB (BF16)", |
| "official_url": "https://huggingface.co/nvidia" |
| }, |
| { |
| "name": "Phi-4", |
| "slug": "phi-4", |
| "convly_url": "https://convly.ai/model/phi-4/", |
| "developer": "Microsoft", |
| "model_type": "LLM (dense)", |
| "modality": "Text → Text", |
| "parameters": "14B", |
| "context_window": "16K", |
| "max_output": null, |
| "license": "MIT (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.07, |
| "output_price": 0.14, |
| "api_providers": "Azure, OpenRouter, Ollama", |
| "vram_q4": "~9 GB", |
| "min_gpu": "RTX 4070 12GB / RTX 3060 12GB", |
| "official_url": "https://huggingface.co/microsoft" |
| }, |
| { |
| "name": "Qwen3 14B", |
| "slug": "qwen3-14b", |
| "convly_url": "https://convly.ai/model/qwen3-14b/", |
| "developer": "Alibaba", |
| "model_type": "LLM (dense)", |
| "modality": "Text → Text", |
| "parameters": "14B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Apache 2.0 (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.12, |
| "output_price": 0.24, |
| "api_providers": "Alibaba, OpenRouter, Ollama", |
| "vram_q4": "~9 GB", |
| "min_gpu": "RTX 4070 12GB (Q4)", |
| "official_url": "https://huggingface.co/Qwen" |
| }, |
| { |
| "name": "Qwen3 235B-A22B", |
| "slug": "qwen3-235b-a22b", |
| "convly_url": "https://convly.ai/model/qwen3-235b-a22b/", |
| "developer": "Alibaba", |
| "model_type": "LLM (MoE)", |
| "modality": "Text → Text", |
| "parameters": "235B total / 22B active (MoE)", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Apache 2.0 (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.45, |
| "output_price": 1.8, |
| "api_providers": "Alibaba, OpenRouter", |
| "vram_q4": "~140 GB", |
| "min_gpu": "Multi-GPU or Mac 192GB", |
| "official_url": "https://huggingface.co/Qwen" |
| }, |
| { |
| "name": "Qwen3 30B-A3B", |
| "slug": "qwen3-30b-a3b", |
| "convly_url": "https://convly.ai/model/qwen3-30b-a3b/", |
| "developer": "Alibaba", |
| "model_type": "LLM (MoE)", |
| "modality": "Text → Text", |
| "parameters": "30B total / 3B active (MoE)", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Apache 2.0 (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.12, |
| "output_price": 0.5, |
| "api_providers": "Alibaba, OpenRouter, Ollama", |
| "vram_q4": "~18 GB", |
| "min_gpu": "RTX 4090 24GB (Q4) — fast, 3B active", |
| "official_url": "https://huggingface.co/Qwen" |
| }, |
| { |
| "name": "Qwen3 32B", |
| "slug": "qwen3-32b", |
| "convly_url": "https://convly.ai/model/qwen3-32b/", |
| "developer": "Alibaba", |
| "model_type": "LLM (dense)", |
| "modality": "Text → Text", |
| "parameters": "32B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Apache 2.0 (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.08, |
| "output_price": 0.28, |
| "api_providers": "Alibaba, OpenRouter, Ollama", |
| "vram_q4": "~20 GB", |
| "min_gpu": "RTX 4090 24GB (Q4)", |
| "official_url": "https://huggingface.co/Qwen" |
| }, |
| { |
| "name": "Qwen3 8B", |
| "slug": "qwen3-8b", |
| "convly_url": "https://convly.ai/model/qwen3-8b/", |
| "developer": "Alibaba", |
| "model_type": "LLM (dense)", |
| "modality": "Text → Text", |
| "parameters": "8B", |
| "context_window": "128K", |
| "max_output": null, |
| "license": "Apache 2.0 (open)", |
| "open_weights": "yes", |
| "release_date": "2025", |
| "input_price": 0.04, |
| "output_price": 0.14, |
| "api_providers": "Alibaba, OpenRouter, Ollama", |
| "vram_q4": "~5 GB", |
| "min_gpu": "RTX 3060 8GB / any 8GB GPU", |
| "official_url": "https://huggingface.co/Qwen" |
| } |
| ] |
| } |
