{ "name": "T-Minus AI Model Catalog", "description": "A source-dated comparison of current AI model pricing, access, licensing, deployment, and workflow fit.", "canonicalPage": "https://www.tminusai.com/models", "verifiedAt": "August 11, 2026", "license": "CC BY 4.0", "methodology": [ "Use provider documentation and official release pages as the factual source.", "Record standard API input and output prices per one million tokens in US dollars only when an official USD price is published.", "Separate provider API prices from self-hosting infrastructure costs and do not label open weights as cost-free.", "Distinguish open source, open weight, and custom-license releases instead of treating them as interchangeable.", "Treat best-for labels as T-Minus AI editorial judgments, not benchmark results.", "Recheck regional plan access and provider limits before a purchase decision." ], "notes": [ "Claude Sonnet 5 introductory API pricing is $2 input and $10 output per million tokens through August 31, 2026. Standard $3/$15 pricing begins September 1, 2026.", "Model access in consumer chat plans varies by provider, region, and plan. Check the linked provider page before purchasing." ], "models": [ { "id": "gpt-5-6-sol", "name": "GPT-5.6 Sol", "provider": "OpenAI", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "ChatGPT Plus", "bestFor": "Frontier coding, knowledge work, and agentic tasks", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://openai.com/index/gpt-5-6/", "verifiedAt": "August 11, 2026" }, { "id": "gpt-5-6-terra", "name": "GPT-5.6 Terra", "provider": "OpenAI", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "ChatGPT Free (limited)", "bestFor": "Balanced everyday execution and coding", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://openai.com/index/gpt-5-6/", "verifiedAt": "August 11, 2026" }, { "id": "gpt-5-6-luna", "name": "GPT-5.6 Luna", "provider": "OpenAI", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "ChatGPT Plus", "bestFor": "Cost-sensitive, high-throughput API work", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://openai.com/index/gpt-5-6/", "verifiedAt": "August 11, 2026" }, { "id": "claude-sonnet-5", "name": "Claude Sonnet 5", "provider": "Anthropic", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "Claude", "bestFor": "Balanced coding, agents, and document-heavy work", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://platform.claude.com/docs/en/about-claude/models/whats-new-sonnet-5", "verifiedAt": "August 11, 2026" }, { "id": "claude-opus-5", "name": "Claude Opus 5", "provider": "Anthropic", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "API and cloud platforms", "bestFor": "Premium reasoning and complex tool work", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://platform.claude.com/docs/en/about-claude/models/overview", "verifiedAt": "August 11, 2026" }, { "id": "claude-fable-5", "name": "Claude Fable 5", "provider": "Anthropic", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "API and cloud platforms", "bestFor": "Most demanding reasoning and long-horizon agents", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://platform.claude.com/docs/en/about-claude/models/introducing-claude-fable-5-and-claude-mythos-5", "verifiedAt": "August 11, 2026" }, { "id": "gemini-3-6-flash", "name": "Gemini 3.6 Flash", "provider": "Google", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "Gemini", "bestFor": "Multimodal and agentic API work", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://ai.google.dev/gemini-api/docs/pricing", "verifiedAt": "August 11, 2026" }, { "id": "gemini-3-5-flash-lite", "name": "Gemini 3.5 Flash-Lite", "provider": "Google", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "Gemini", "bestFor": "High-volume, low-cost API tasks", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://ai.google.dev/gemini-api/docs/pricing", "verifiedAt": "August 11, 2026" }, { "id": "grok-4-5", "name": "Grok 4.5", "provider": "xAI", "availability": "Hosted", "licenseType": "Provider terms", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "Grok", "bestFor": "General work and API experimentation", "localUse": "Provider-hosted access", "detailPage": null, "officialSource": "https://x.ai/news/grok-4-5", "verifiedAt": "August 11, 2026" }, { "id": "kimi-k3", "name": "Kimi K3", "provider": "Moonshot AI", "availability": "Open-weight", "licenseType": "Kimi K3 License", "modalities": [ "Text", "Image" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "See official source", "chatAccess": "Kimi API and certified inference partners", "bestFor": "Long-horizon coding agents", "localUse": "Official documentation recommends vLLM, SGLang, or TokenSpeed. At 2.8T total parameters, this is infrastructure-scale deployment, not a realistic local laptop model.", "detailPage": "/models/kimi-k3", "officialSource": "https://github.com/MoonshotAI/Kimi-K3", "verifiedAt": "August 11, 2026" }, { "id": "minimax-m3", "name": "MiniMax M3", "provider": "MiniMax", "availability": "Open-weight", "licenseType": "Open-weight release, verify model-card terms", "modalities": [ "Text", "Image", "Video input" ], "inputUsdPerMillion": 0.3, "outputUsdPerMillion": 1.2, "pricingBasis": "USD per 1M tokens", "chatAccess": "MiniMax API and Token Plan", "bestFor": "Agentic coding", "localUse": "MiniMax documents SGLang, vLLM, Transformers, KTransformers, and Unsloth. Its 428B total parameter footprint makes managed access the practical starting point for most builders.", "detailPage": "/models/minimax-m3", "officialSource": "https://github.com/MiniMax-AI/MiniMax-M3", "verifiedAt": "August 11, 2026" }, { "id": "qwen-3-6", "name": "Qwen3.6", "provider": "Qwen / Alibaba", "availability": "Open source", "licenseType": "Apache 2.0", "modalities": [ "Text", "Vision variants" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "USD per 1M tokens", "chatAccess": "Qwen Studio and Alibaba Cloud Model Studio", "bestFor": "Local coding assistance", "localUse": "Qwen documents Transformers, llama.cpp, MLX and MLX-VLM on Apple Silicon, SGLang, and vLLM. Smaller and quantized Qwen variants are among the more approachable current open models for local development.", "detailPage": "/models/qwen-3-6", "officialSource": "https://github.com/QwenLM/Qwen3.6", "verifiedAt": "August 11, 2026" }, { "id": "glm-5-2", "name": "GLM-5.2", "provider": "Z.ai", "availability": "Open source", "licenseType": "MIT", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "USD per 1M tokens", "chatAccess": "Z.ai general API and GLM Coding Plan", "bestFor": "Long-horizon coding agents", "localUse": "Z.ai documents Transformers, vLLM, SGLang, xLLM, and KTransformers. The 744B total parameter footprint means local use is a data-center or serious multi-GPU decision, even though the model is openly available.", "detailPage": "/models/glm-5-2", "officialSource": "https://z.ai/blog/glm-5.2", "verifiedAt": "August 11, 2026" }, { "id": "deepseek-v4", "name": "DeepSeek V4", "provider": "DeepSeek", "availability": "Hosted API with published model ecosystem", "licenseType": "Verify current weight terms", "modalities": [ "Text" ], "inputUsdPerMillion": 0.14, "outputUsdPerMillion": 0.28, "pricingBasis": "USD per 1M tokens", "chatAccess": "DeepSeek API", "bestFor": "Cost-sensitive API workflows", "localUse": "Treat the exact V4 deployment path as a provider-specific verification item. The current official source cited here is the hosted API documentation, not a self-hosting instruction set.", "detailPage": "/models/deepseek-v4", "officialSource": "https://api-docs.deepseek.com/quick_start/pricing/", "verifiedAt": "August 11, 2026" }, { "id": "mistral-small-4", "name": "Mistral Small 4", "provider": "Mistral AI", "availability": "Open source", "licenseType": "Apache 2.0", "modalities": [ "Text", "Image" ], "inputUsdPerMillion": 0.15, "outputUsdPerMillion": 0.6, "pricingBasis": "USD per 1M tokens", "chatAccess": "Mistral API and Mistral AI Studio", "bestFor": "Multimodal document analysis", "localUse": "Mistral documents support through vLLM, llama.cpp, SGLang, and Transformers. Its official minimum deployment examples use multi-GPU hardware, so a smaller quantization is more realistic for an individual local setup.", "detailPage": "/models/mistral-small-4", "officialSource": "https://mistral.ai/news/mistral-small-4/", "verifiedAt": "August 11, 2026" }, { "id": "mistral-large-3", "name": "Mistral Large 3", "provider": "Mistral AI", "availability": "Open source", "licenseType": "Apache 2.0", "modalities": [ "Text", "Image understanding" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "USD per 1M tokens", "chatAccess": "Mistral AI Studio and selected cloud providers", "bestFor": "Enterprise custom models", "localUse": "Mistral documents optimized NVFP4 serving on Blackwell NVL72 systems and multi-GPU A100 or H100 nodes. This is a cloud or data-center open model, not a typical workstation deployment.", "detailPage": "/models/mistral-large-3", "officialSource": "https://mistral.ai/news/mistral-3/", "verifiedAt": "August 11, 2026" }, { "id": "gemma-4", "name": "Gemma 4", "provider": "Google", "availability": "Open model family", "licenseType": "Gemma 4 license", "modalities": [ "Text", "Unified and specialist variants" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "USD per 1M tokens", "chatAccess": "Downloadable model family and Google AI ecosystem", "bestFor": "Right-sized open-model experiments", "localUse": "Gemma spans small and larger variants, so the practical local path depends on the checkpoint and quantization. Treat it as a family selection problem rather than a one-model hardware estimate.", "detailPage": "/models/gemma-4", "officialSource": "https://ai.google.dev/gemma/docs/releases?hl=en", "verifiedAt": "August 11, 2026" }, { "id": "gpt-oss-120b", "name": "gpt-oss-120b", "provider": "OpenAI", "availability": "Open-weight", "licenseType": "Apache 2.0", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "USD per 1M tokens", "chatAccess": "Self-hosted or third-party hosting providers", "bestFor": "Private-cloud reasoning services", "localUse": "OpenAI states that the native MXFP4 120B release can run within 80GB of memory. It is a realistic single high-memory GPU or hosted-inference option, not a typical consumer laptop model.", "detailPage": "/models/gpt-oss-120b", "officialSource": "https://openai.com/index/introducing-gpt-oss/", "verifiedAt": "August 11, 2026" }, { "id": "gpt-oss-20b", "name": "gpt-oss-20b", "provider": "OpenAI", "availability": "Open-weight", "licenseType": "Apache 2.0", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "USD per 1M tokens", "chatAccess": "Self-hosted or third-party hosting providers", "bestFor": "Local text reasoning", "localUse": "OpenAI states that the native MXFP4 gpt-oss-20b release can run with 16GB of memory, which makes it one of the more approachable current reasoning models for local or edge deployment.", "detailPage": "/models/gpt-oss-20b", "officialSource": "https://openai.com/index/introducing-gpt-oss/", "verifiedAt": "August 11, 2026" }, { "id": "nemotron-3-super", "name": "Nemotron 3 Super", "provider": "NVIDIA", "availability": "Open-weight", "licenseType": "NVIDIA Open Model License", "modalities": [ "Text" ], "inputUsdPerMillion": null, "outputUsdPerMillion": null, "pricingBasis": "USD per 1M tokens", "chatAccess": "Downloadable weights, NVIDIA services, and selected hosting providers", "bestFor": "Enterprise agent systems", "localUse": "Nemotron 3 Super is optimized around NVIDIA accelerated infrastructure and NVFP4-oriented deployment. It is suitable for teams already operating NVIDIA hardware or cloud GPU capacity, not a generic laptop target.", "detailPage": "/models/nemotron-3-super", "officialSource": "https://developer.nvidia.com/blog/introducing-nemotron-3-super-an-open-hybrid-mamba-transformer-moe-for-agentic-reasoning/", "verifiedAt": "August 11, 2026" } ] }