{
  "site": "https://littleagentclub.com",
  "evidence": "Provider descriptions, not our own hands-on findings.",
  "models": [
    {
      "id": "nemotron-3-5-lightning",
      "name": "Nemotron 3.5 Lightning",
      "kind": "model",
      "url": "https://littleagentclub.com/models#nemotron-3-5-lightning",
      "description": "Explore a model for tool use and agent workflows on your own hardware or through an API.",
      "access": "Downloadable weights · OpenMDW 1.1",
      "notes": "30B total parameters, 3B active. NVIDIA documents both local and server setups.",
      "checkedOn": "17 October 2026",
      "sources": [
        {
          "label": "NVIDIA model card",
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4"
        },
        {
          "label": "Read the license",
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4/blob/main/LICENSE"
        },
        {
          "label": "OpenRouter provider prices",
          "url": "https://openrouter.ai/nvidia/nemotron-3.5-lightning"
        }
      ]
    },
    {
      "id": "llama-4-scout",
      "name": "Llama 4 Scout",
      "kind": "model",
      "url": "https://littleagentclub.com/models#llama-4-scout",
      "description": "Build a text-and-image assistant with a downloadable member of the Llama family.",
      "access": "Gated weights · Llama 4 Community License",
      "notes": "17B active parameters does not mean a 17B download: Scout has 109B parameters in total.",
      "checkedOn": "17 October 2026",
      "sources": [
        {
          "label": "Read the license and terms",
          "url": "https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct"
        },
        {
          "label": "OpenRouter provider prices",
          "url": "https://openrouter.ai/meta-llama/llama-4-scout"
        }
      ]
    },
    {
      "id": "deepseek-v4-1-flash",
      "name": "DeepSeek V4.1 Flash",
      "kind": "model",
      "url": "https://littleagentclub.com/models#deepseek-v4-1-flash",
      "description": "Explore image understanding, reasoning and tool use through a model you can download or call by API.",
      "access": "Downloadable weights · MIT",
      "notes": "A large model despite “Flash”: 552B backbone parameters, plus a 196B-parameter Engram component.",
      "checkedOn": "2026-10-21",
      "sources": [
        {
          "label": "DeepSeek model card",
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash"
        },
        {
          "label": "DeepSeek inference guide",
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash/blob/main/inference/README.md"
        },
        {
          "label": "Read the license",
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash/blob/main/LICENSE"
        },
        {
          "label": "DeepSeek API pricing",
          "url": "https://api-docs.deepseek.com/quick_start/pricing/"
        }
      ]
    },
    {
      "id": "glm-5-3",
      "name": "GLM 5.3",
      "kind": "model",
      "url": "https://littleagentclub.com/models#glm-5-3",
      "description": "Explore a released text-only coding and reasoning model for multi-step software projects.",
      "access": "Downloadable weights · GLM-5.3 License",
      "notes": "The published FP8 download is about 756 GB. This is a server-scale model.",
      "checkedOn": "2026-10-21",
      "sources": [
        {
          "label": "Z.ai model card and files",
          "url": "https://huggingface.co/zai-org/GLM-5.3"
        },
        {
          "label": "SGLang serving guide",
          "url": "https://docs.sglang.io/cookbook/autoregressive/GLM/GLM-5.3"
        },
        {
          "label": "Read the license",
          "url": "https://huggingface.co/zai-org/GLM-5.3/blob/main/LICENSE"
        },
        {
          "label": "Z.ai API pricing",
          "url": "https://docs.z.ai/guides/overview/pricing"
        }
      ]
    },
    {
      "id": "qwen-3-6",
      "name": "Qwen3.6-27B",
      "kind": "model",
      "url": "https://littleagentclub.com/models#qwen-3-6",
      "description": "For builders who want to work through code, documents and images using a model they can download or call through a host.",
      "access": "Downloadable weights · Apache 2.0",
      "notes": "This is the exact 27B release, not the newer Qwen models or the Qwen app. Its native context is 262,144 tokens; extending it needs a suitable setup. Context capacity is not a free usage allowance.",
      "checkedOn": "21 October 2026; company sources checked 22 October 2026",
      "sources": [
        {
          "label": "Pinned model card",
          "url": "https://huggingface.co/Qwen/Qwen3.6-27B/blob/6a9e13bd6fc8f0983b9b99948120bc37f49c13e9/README.md"
        },
        {
          "label": "Official serving examples",
          "url": "https://huggingface.co/Qwen/Qwen3.6-27B/blob/6a9e13bd6fc8f0983b9b99948120bc37f49c13e9/README.md#quickstart"
        },
        {
          "label": "Exact checkpoint license",
          "url": "https://huggingface.co/Qwen/Qwen3.6-27B/blob/6a9e13bd6fc8f0983b9b99948120bc37f49c13e9/LICENSE"
        },
        {
          "label": "Direct regional model information",
          "url": "https://www.alibabacloud.com/help/en/model-studio/qwen3-6-27b"
        }
      ]
    },
    {
      "id": "muse-spark-open",
      "name": "Muse Spark Open",
      "kind": "model",
      "url": "https://littleagentclub.com/models#muse-spark-open",
      "description": "For developers building coding and visual tools who want downloadable weights and can arrange substantial computing power or hosted access.",
      "access": "Gated weights · custom Muse Model License",
      "notes": "Meta describes 816B total parameters with 42B active at a time and a 1M-token context. Active parameters do not describe the full download. Spark Open is different from the closed Spark 1.3 and Wonder models.",
      "checkedOn": "21 October 2026; company and routing sources checked 22 October 2026",
      "sources": [
        {
          "label": "Exact model release",
          "url": "https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B/blob/0fb02478b02565b2ce7fc57e6051c91172774ac9/README.md"
        },
        {
          "label": "Release and documented local example",
          "url": "https://research.meta.ai/blog/muse-spark-open"
        },
        {
          "label": "Release license and access request",
          "url": "https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B"
        },
        {
          "label": "Model rates and provider choices",
          "url": "https://openrouter.ai/meta/muse-spark-open"
        }
      ]
    },
    {
      "id": "muse-wonder-1-0",
      "name": "Muse Wonder 1.0",
      "kind": "model",
      "url": "https://littleagentclub.com/models#muse-wonder-1-0",
      "description": "For builders who want a model to work through larger coding projects, documents and tool-guided tasks through an API.",
      "access": "Hosted preview · no downloadable weights",
      "notes": "The direct API model is muse-wonder-1.0 with 1,048,576-token context. Wonder Full is a different variant available through eligible Muse Code plans. Do not transfer its results or access to this API model.",
      "checkedOn": "21 October 2026; company sources checked 22 October 2026",
      "sources": [
        {
          "label": "Current model identity and preview status",
          "url": "https://dev.meta.ai/docs/overview"
        },
        {
          "label": "Account setup and payment prerequisites",
          "url": "https://dev.meta.ai/help/accounts-and-login/sign-up"
        },
        {
          "label": "Direct rates and search charges",
          "url": "https://dev.meta.ai/docs/pricing-rate-limits"
        }
      ]
    },
    {
      "id": "beam-1-0",
      "name": "Beam 1.0",
      "kind": "model",
      "url": "https://littleagentclub.com/models#beam-1-0",
      "description": "For builders who want a downloadable model for code and multi-step tasks, or prefer a separate hosted service.",
      "access": "Downloadable weights · Apache 2.0",
      "notes": "Text input and output, with a published 1M-token context. The full model has about 501B parameters; about 23B are active at once. That does not make it a small laptop download.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Exact Beam release",
          "url": "https://reflection.ai/beam-model-card"
        },
        {
          "label": "Download and reference hardware",
          "url": "https://developers.reflection.ai/serving-weights"
        },
        {
          "label": "Pinned weight license",
          "url": "https://huggingface.co/reflection/Beam-1.0-501B-A23B/blob/78f8f654d40a40919b8e9954b40dd879ecb9a692/LICENSE"
        },
        {
          "label": "Model routes and current token rates",
          "url": "https://openrouter.ai/reflection/beam-1.0-501b-a23b"
        }
      ]
    },
    {
      "id": "kolibri-1",
      "name": "Kolibri-1",
      "kind": "model",
      "url": "https://littleagentclub.com/models#kolibri-1",
      "description": "For builders working with German or English documents who can arrange their own model hosting.",
      "access": "Downloadable weights · Apache 2.0",
      "notes": "A text model with about 78B total parameters and about 3B active at once. The published context reaches 1,048,576 tokens; the provider recommends 262,144 for efficient serving and complex work. These are technical limits, not free usage allowances.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Exact Kolibri release and limits",
          "url": "https://aleph-alpha.com/en/kolibri/"
        },
        {
          "label": "Official serving guide",
          "url": "https://huggingface.co/Aleph-Alpha/Kolibri-1"
        },
        {
          "label": "Pinned weight-license scope",
          "url": "https://huggingface.co/Aleph-Alpha/Kolibri-1/blob/5bb7075ba22db733e3834b1c71d2245537cc14ed/README.md#license-and-terms"
        },
        {
          "label": "Current API interest and access page",
          "url": "https://aleph-alpha.com/en/signup-kolibri-api/"
        }
      ]
    },
    {
      "id": "minimax-m31-flash",
      "name": "MiniMax M3.1 Flash",
      "kind": "model",
      "url": "https://littleagentclub.com/models#minimax-m31-flash",
      "description": "For builders who want a model for coding, documents and tool-guided work, with text, image and video input.",
      "access": "Hosted model · weights promised later",
      "notes": "The current API guide lists a 1M-token context. Thinking is always on and defaults to the highest effort; lowering effort can reduce usage. The 19 October launch says weights will be released later. No downloadable M3.1 checkpoint or weight license was confirmed.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Current model identity and capabilities",
          "url": "https://platform.minimax.io/docs/guides/text-generation"
        },
        {
          "label": "Account keys and shared usage",
          "url": "https://platform.minimax.io/docs/m-plan/faq"
        },
        {
          "label": "Current release and weight status",
          "url": "https://www.minimax.io/blog/minimax-m31f"
        },
        {
          "label": "Direct model rates",
          "url": "https://platform.minimax.io/docs/pricing/overview"
        }
      ]
    },
    {
      "id": "step-5-preview",
      "name": "Step 5 Preview",
      "kind": "model",
      "url": "https://littleagentclub.com/models#step-5-preview",
      "description": "For builders working through code, long documents and images who can arrange an API connection or substantial model hosting.",
      "access": "Downloadable weights · custom StepFun license",
      "notes": "The official FP8 release has 28 weight files totaling about 644 GB. About 27B parameters are active at once; the whole model is about 600B. The hosted API lists 1M context and 64K maximum output. The download card lists different output limits, so check your chosen deployment.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Pinned model release and deployment card",
          "url": "https://huggingface.co/stepfun-ai/Step-5-Preview/blob/364b3b03fed2477aa0953fe855a2c4c8c48ee3cc/README.md"
        },
        {
          "label": "Official serving configurations",
          "url": "https://huggingface.co/stepfun-ai/Step-5-Preview/blob/364b3b03fed2477aa0953fe855a2c4c8c48ee3cc/README.md#deployment"
        },
        {
          "label": "Exact release license",
          "url": "https://huggingface.co/stepfun-ai/Step-5-Preview/blob/364b3b03fed2477aa0953fe855a2c4c8c48ee3cc/LICENSE.txt"
        },
        {
          "label": "Direct rates and additional charges",
          "url": "https://platform.stepfun.ai/docs/en/guides/pricing/details"
        }
      ]
    },
    {
      "id": "mistral-large-4",
      "name": "Mistral Large 4",
      "kind": "model",
      "url": "https://littleagentclub.com/models#mistral-large-4",
      "description": "For builders testing a model that reads documents and images, writes code and works with tools.",
      "access": "Hosted API · public preview",
      "notes": "The model page lists a 1M-token context. This is a capacity limit, not a free allowance. The API is available as a preview; downloadable weights and their license still say Coming soon.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Exact model and current release status",
          "url": "https://docs.mistral.ai/models/mistral-large-4"
        },
        {
          "label": "Release announcement",
          "url": "https://mistral.ai/news/mistral-large-4/"
        },
        {
          "label": "Preview terms, section 4.3",
          "url": "https://legal.mistral.ai/terms/commercial-terms-of-service/"
        },
        {
          "label": "Direct pricing and currency selector",
          "url": "https://docs.mistral.ai/inference/pricing"
        }
      ]
    },
    {
      "id": "jev-1-13",
      "name": "Jev 1.13",
      "kind": "model",
      "url": "https://littleagentclub.com/models#jev-1-13",
      "description": "For developers who want software to sort requests, pick from options or score text against clear rules.",
      "access": "Hosted decision model · stable release",
      "notes": "Jev returns typed decisions and probabilities rather than chat replies. The stable jev-latest alias currently points to 1.13.0; the newer jev-preview alias points to 1.14.0. One model entry covers this distinction.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Exact model pricing and context rules",
          "url": "https://docs.typesafe.ai/models"
        },
        {
          "label": "Request shape and response fields",
          "url": "https://docs.typesafe.ai/api/system-one"
        },
        {
          "label": "Known limitations for this version",
          "url": "https://docs.typesafe.ai/model-jaggedness/jev-1.13"
        }
      ]
    },
    {
      "id": "gpt-6-sol",
      "name": "GPT-6 Sol",
      "kind": "model",
      "url": "https://littleagentclub.com/models#gpt-6-sol",
      "description": "Explore a hosted model for coding and multi-step tasks. It is a different release from GPT-6.1 Sol.",
      "access": "Hosted API · usage billing",
      "notes": "Text and image input, text output. The model guide lists a 1,050,000-token context and 128,000-token output ceiling. These are request limits, not free-use quotas.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Exact Sol pricing and context rules",
          "url": "https://developers.openai.com/api/docs/models/gpt-6-sol"
        }
      ]
    },
    {
      "id": "gpt-6-luna",
      "name": "GPT-6 Luna",
      "kind": "model",
      "url": "https://littleagentclub.com/models#gpt-6-luna",
      "description": "Explore a hosted model for short, repeated work such as reading and sorting information.",
      "access": "Hosted API · usage billing",
      "notes": "Text and image input, text output. The model guide lists a 1,050,000-token context and 128,000-token output ceiling. The amount an account can send also depends on its rate limits.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Exact Luna pricing and API compatibility",
          "url": "https://developers.openai.com/api/docs/models/gpt-6-luna"
        }
      ]
    },
    {
      "id": "claude-opus-5-5",
      "name": "Claude Opus 5.5",
      "kind": "model",
      "url": "https://littleagentclub.com/models#claude-opus-5-5",
      "description": "Explore a hosted model for larger software changes and work with long documents. Model access and an assistant subscription are different offers.",
      "access": "Hosted API · usage billing",
      "notes": "Text and image input, text output. Standard requests support a 1M-token context and up to 128K output tokens. These are technical ceilings, not included monthly usage.",
      "checkedOn": "21 October 2026",
      "sources": [
        {
          "label": "Provider access documentation",
          "url": "https://platform.claude.com/docs/en/models/opus-5-5/overview"
        },
        {
          "label": "Token, cache, Fast, Batch and regional pricing",
          "url": "https://platform.claude.com/docs/en/about-claude/pricing"
        }
      ]
    },
    {
      "id": "flint",
      "name": "Flint",
      "kind": "model",
      "url": "https://littleagentclub.com/models#flint",
      "description": "A local coding model for fixes, features and tests.",
      "access": "Downloadable weights · Apache 2.0",
      "notes": "The model card lists 39.8B total parameters and about 2.5B active per token. It works with text and tools. A coding app supplies the tools and permissions.",
      "checkedOn": "22 October 2026",
      "sources": [
        {
          "label": "Exact Flint model card",
          "url": "https://huggingface.co/RecursiveSI/flint/blob/2a247be0c0aec895ccb0ac7ca92229ec6df37288/README.md"
        },
        {
          "label": "rec setup and hardware requirements",
          "url": "https://docs.recursive.com/edge/rec"
        },
        {
          "label": "Checkpoint license",
          "url": "https://huggingface.co/RecursiveSI/flint/blob/2a247be0c0aec895ccb0ac7ca92229ec6df37288/LICENSE"
        },
        {
          "label": "OpenCode free access and local setup",
          "url": "https://docs.recursive.com/guides/coding-agents"
        }
      ]
    },
    {
      "id": "muse-glimmer",
      "name": "Muse Glimmer",
      "kind": "model",
      "url": "https://littleagentclub.com/models#muse-glimmer",
      "description": "Explore an agent model on a capable computer, with tool use and image understanding.",
      "access": "Downloadable weights · Apache 2.0",
      "notes": "The 30B model still needs substantial memory; a quantized version fits a 24–32 GB envelope.",
      "checkedOn": "17 October 2026",
      "sources": [
        {
          "label": "Meta research",
          "url": "https://research.meta.ai/blog/introducing-muse-glimmer-open-agentic-model"
        }
      ]
    },
    {
      "id": "gemma-4",
      "name": "Gemma 4",
      "kind": "model",
      "url": "https://littleagentclub.com/models#gemma-4",
      "description": "Find a size that fits your project, from edge devices to larger local or hosted setups.",
      "access": "Downloadable weights · Apache 2.0",
      "notes": "Choose a specific variant before estimating device requirements or running cost.",
      "checkedOn": "17 October 2026",
      "sources": [
        {
          "label": "Google Open Source",
          "url": "https://opensource.googleblog.com/2026/03/gemma-4-expanding-the-gemmaverse-with-apache-20.html"
        }
      ]
    },
    {
      "id": "olmo-3",
      "name": "Olmo 3.1 Instruct 32B",
      "kind": "model",
      "url": "https://littleagentclub.com/models#olmo-3",
      "description": "Study, adapt and reproduce a model with unusually complete training materials.",
      "access": "Weights, code & training data · Apache 2.0",
      "notes": "This 32B checkpoint includes a quantized inference example. Check the framework and memory needs for your setup.",
      "checkedOn": "17 October 2026",
      "sources": [
        {
          "label": "Ai2 model card",
          "url": "https://huggingface.co/allenai/Olmo-3.1-32B-Instruct"
        }
      ]
    },
    {
      "id": "mistral-small-4",
      "name": "Mistral Small 4",
      "kind": "model",
      "url": "https://littleagentclub.com/models#mistral-small-4",
      "description": "Explore multimodal, reasoning and coding work in one released model.",
      "access": "Downloadable weights · Apache 2.0",
      "notes": "Mistral’s published serving setup uses server GPUs. A smaller active parameter count does not describe the full download.",
      "checkedOn": "17 October 2026",
      "sources": [
        {
          "label": "Mistral release",
          "url": "https://mistral.ai/news/mistral-small-4/"
        }
      ]
    }
  ]
}