AI model finder · verified source catalogs
Find an AI model that fits your work
Search in plain English, then refine the full catalog with advanced filters.
AI-powered model search
Describe what you need
Detection runs in your browser. Searches and the selected hardware profile are logged to improve recommendations; no email address is collected here. Privacy details
Model results
723 matches · page 1 of 35 · 21 per page
rlmath-agentic-grpo-checkpoints
rlmath-agentic-grpo-checkpoints is a Hugging Face repository presented as a reinforcement learning model for agentic applications. The supplied record classifies its capabilities as agentic and lists agents and agentic workflows as practical use cases, so it may be relevant for developers investigating reinforcement-learning checkpoints intended for autonomous or multi-step agent behavior. The supplied material does not identify a specific model task beyond the repository’s general reinforcement-learning description, and it provides no verified information about supported input or output modalities. Parameter count, context window, quantization format, inference speed, pricing, benchmark results, and runtime requirements are not available in the supplied material. This record is licensed under Apache-2.0; that license applies to this repository and should be reviewed alongside any applicable files or dependencies before deployment. Because the excerpt contains only the repository metadata and does not include a model card, training details, checkpoint naming scheme, evaluation results, or usage instructions, developers should treat architectural, performance, compatibility, and safety characteristics as unverified.
- Model size
- Size not listed
- Context
- Not listed
inkling-2-small
inkling-2-small is a 594-billion-parameter Mixture-of-Experts model with 17 billion parameters active, designed for agentic coding and knowledge work. The supplied material lists image input and tool-calling capability, making it relevant to coding workflows, software agents, agentic automation, and vision-enabled tasks. Its advertised context window is up to 1 million tokens; no separate runtime context configuration or default is provided, so deployment settings remain unverified. The source does not provide inference-speed benchmarks, pricing, a model task classification, license, or quantization details. Consequently, this record cannot establish whether the Ollama entry is quantized or describe the applicable license for redistribution or commercial use. The listed modality is image, while the summary specifically describes image input; developers should verify any additional input or output modalities and the exact tool-calling interface in their target runtime. The very large total parameter count, despite the lower active-parameter figure, may affect hosting requirements, but hardware, memory, and performance requirements are not available in the supplied material.
- Model size
- 594B parameters
- Context
- Not listed
dolphin3:8b
dolphin3:8b is an instruct-tuned, general-purpose local model based on Dolphin 3.0 Llama 3.1 8B. The supplied model description specifically positions it for coding, math, agentic workflows, function calling, and general use, making it a candidate for local developer assistants, autonomous task orchestration, and tool-integrated applications. Its advertised focus suggests practical value when you need one model to cover several workloads rather than a narrowly specialized system. The source excerpt does not provide a context-window size, modality list, detailed capability specifications, inference benchmark, pricing, or licensing information, so those factors should be verified before deployment. It also does not establish a default runtime context configuration or production performance. The model is identified as 8B, and the supplied record reports 8,000,000,000 parameters. No tokens-per-second result is available in the supplied material. Because the page describes broad intended uses but gives limited technical detail, evaluate it with your own coding, reasoning, function-calling, and agent workflow tests before relying on it in production.
- Model size
- 8B parameters
- Context
- Not listed
qwen3-coder-next
qwen3-coder-next is a text-only, coding-focused language model intended for developers building agentic software workflows. The supplied material describes it as coming from Alibaba’s Qwen team and optimized for agentic coding and local development. Its listed tool-calling capability makes it relevant for coding agents that need to interact with development tools, execute structured actions, or coordinate multistep programming tasks. In practical terms, it is a reasonable model to evaluate for code generation, code assistance, repository-oriented automation, and agentic development workflows through Ollama. The available information is limited: it does not provide a context-window size, parameter count, inference-speed benchmark, pricing, or detailed guidance about supported programming languages, hardware requirements, evaluation results, or production reliability. The supplied material also contains no licensing information, so licensing compatibility should be verified before redistribution, commercial deployment, or embedding the model in a product. Because no advertised context window or runtime configuration is provided, long-context suitability cannot be assessed from this material. Developers should validate tool-calling behavior, code accuracy, latency, and resource usage in their own Ollama environment before relying on it for unattended or high-impact automation.
- Model size
- Size not listed
- Context
- Not listed
qwen3-coder:30b
qwen3-coder:30b is a coding-focused model intended for agentic workflows and tool-enabled development. The supplied Ollama material describes it as offering long-context capability for coding and agentic tasks, while the model record lists tool-calling as a capability. This makes it a reasonable candidate for software agents that need to generate or modify code, coordinate multi-step development actions, and interact with tools. Its advertised context window is not available in the supplied material, so the practical amount of code or project context it can handle cannot be determined here; no runtime context configuration is provided either. No modalities are listed, so support for image, audio, or other non-text inputs should not be assumed. Inference speed and pricing are also not available in the supplied material. The source excerpt does not provide licensing terms, so review the model’s applicable license and usage conditions before deploying it commercially or redistributing it. The page identifies the model as qwen3-coder:30b and provides a concise description, but does not include benchmark results, quality comparisons, hardware requirements, or detailed safety guidance. Evaluate it with your own codebase and tool-calling workflows before relying on it in production.
- Model size
- 30B parameters
- Context
- Not listed
qwen3-coder:480b
qwen3-coder:480b is intended for coding and agentic development workflows, with tool-calling listed as a capability. The supplied description presents it as one of Alibaba’s performant long-context models for agentic and coding tasks, making it relevant for code generation, code modification, and software agents that may invoke external tools. The model is listed with 480 billion parameters. Its advertised context-window size is not available in the supplied material, so the “long context” description should not be treated as a specific token limit or runtime configuration. No input, output, or other modality is documented here, and no inference-speed benchmark is provided. The source is an Ollama model page for this exact model, but the supplied material does not provide licensing terms, usage restrictions, pricing, hardware requirements, or comparative benchmark results. Developers should therefore validate deployment feasibility, tool-integration behavior, context limits, and licensing before adopting it in production. Its strongest documented fit is coding and agentic use rather than a broader multimodal workload.
- Model size
- 480B parameters
- Context
- Not listed
gpt-oss:120b
gpt-oss:120b is an open-weight model presented for powerful reasoning, agentic tasks, and versatile developer use cases. It is a practical candidate for autonomous agents, workflows that require multi-step reasoning, and applications that need tool-calling. The supplied model record also identifies reasoning, agents, and agentic work as relevant use cases, while tool-calling is listed as a capability. No input or output modalities are listed in the supplied material, so suitability for vision, audio, image, or video workloads cannot be established here. The model is identified as having 120 billion parameters, but the supplied material does not provide a context-window value, inference-speed benchmark, or pricing information. It is hosted through Ollama at the supplied model page, which provides an installation and distribution route, but runtime configuration details are not included in the excerpt. Although the description calls it open-weight, specific license terms and usage restrictions are not available in the supplied material; review the applicable model license before deploying or redistributing it. This makes it best suited for developers prioritizing reasoning and agentic functionality who can validate operational requirements independently.
- Model size
- 120B parameters
- Context
- Not listed
nemotron-3-nano:4b
nemotron-3-nano:4b is a 4-billion-parameter model positioned for efficient, open, and intelligent agentic applications. Its documented capability is tool-calling, making it relevant for agents that need to select or invoke external tools as part of a workflow. The supplied model record identifies agents and agentic workloads as use cases, with tool-calling also listed as a capability. No modalities are listed in the supplied material, so modality-specific support cannot be confirmed. The advertised context window, inference speed, and pricing are not available in the supplied material, which limits direct assessment of long-context tasks, throughput, and operating cost. The source page describes this as an updated 4B parameter-count model, but provides no benchmark results or comparative performance data. Licensing terms and usage restrictions are also not available in the supplied material, so review the applicable Ollama or model documentation before deployment, especially for commercial or redistributed applications. This model is best considered when a compact agentic model with tool-calling is the primary requirement and missing operational specifications can be validated separately.
- Model size
- 4B parameters
- Context
- Not listed
gpt-oss:20b
gpt-oss:20b is an open-weight model aimed at developers building reasoning-driven applications and agentic workflows. The supplied material advertises powerful reasoning, agentic tasks, and versatile developer use cases, while the model record lists agents, agentic work, reasoning, and tool-calling as relevant capabilities or uses. This makes it a candidate for applications that need multi-step reasoning, autonomous task execution, or integration with external tools. No modalities are listed in the supplied material, so support for vision, audio, image, or video input and output is not established here. The advertised context window is also not available in the supplied material, and no inference speed benchmark or tokens-per-second figure is provided. The model record identifies the size as 20,000,000,000 parameters. Because the source describes it as open-weight but does not provide licensing terms in the supplied excerpt, review the applicable license and usage conditions before deployment. Pricing, hardware requirements, quality benchmarks, and safety characteristics are likewise not available in the supplied material, so evaluate those factors independently for your intended workload.
- Model size
- 20B parameters
- Context
- Not listed
nemotron-3-nano:30b
nemotron-3-nano:30b is a 30-billion-parameter Ollama model aimed at agentic applications and supports tool-calling. It may be a practical option for developers building agents that need to invoke external tools, coordinate multi-step tasks, or integrate model responses into automated workflows. The supplied material describes Nemotron-3-Nano as an efficient, open, and intelligent agentic model, but it does not provide benchmark results, inference-speed measurements, or quality comparisons for this exact 30b variant. No advertised context-window size is available in the supplied material, so deployment limits should be verified before selecting it for long-context workloads. The page’s description also mentions an updated 4B parameter-count model, which introduces variant ambiguity despite this record identifying the requested model as 30b; confirm the intended artifact and tag before deployment. No modalities are listed, so image, audio, or video input and output should not be assumed. Pricing and licensing terms are not available in the supplied material. Overall, consider it for tool-using agents, but validate context capacity, runtime behavior, licensing, and the exact variant in your target Ollama environment.
- Model size
- 30B parameters
- Context
- Not listed
ornith:35b
ornith:35b is a 35-billion-parameter open-source model intended primarily for agentic coding workflows. Its advertised focus is a self-improving model family for agentic coding, making it relevant for software development, coding assistance, and autonomous or semi-autonomous agent tasks. The supplied record also identifies tool-calling as a capability, which may support workflows where an agent interacts with external tools or services. No modalities are listed in the supplied material, so support for vision, audio, image, or video inputs and outputs is not established here. The context window, inference speed, and pricing are not available in the supplied material. The model is presented as open source, but no specific license, usage restrictions, safety policy, or deployment requirements are provided, so those details should be verified before commercial or production use. The available information does not include benchmark results, quality comparisons, hardware requirements, quantization details, or runtime defaults. Developers should therefore evaluate its coding accuracy, tool-use reliability, latency, resource consumption, and behavior on their own workloads before adopting it.
- Model size
- 35B parameters
- Context
- Not listed
ripple-preview:65b
ripple-preview:65b is a 65B dense model aimed at agentic coding and workflows that can use tools. Its advertised context window is up to 1 million tokens, which may suit large codebases, long-running coding sessions, and agents that need to retain substantial task context. The supplied record lists coding, agents, and agentic work as use cases, and identifies tool-calling as a capability. No input or output modalities are listed, so support for vision, audio, or other non-text modalities is not established here. The model’s model task, runtime context configuration, quantization details, and license are not available in the supplied material; the 1-million-token figure should therefore be treated as the advertised context limit rather than a confirmed default runtime setting. This record represents the Ollama model entry for ripple-preview:65b, but it does not provide inference-speed benchmarks, quality evaluations, pricing, or deployment requirements. The parameter count is explicitly given as 65,000,000,000. Because quantization is unspecified, the material does not establish whether this Ollama entry is a quantized upload or an unquantized model.
- Model size
- 65B parameters
- Context
- Not listed
gemma4:26b
gemma4:26b is a multimodal model intended for reasoning, coding, and agentic workflows. The supplied material describes Gemma 4 models as targeting frontier-level performance across sizes and highlights reasoning, agentic workflows, coding, and multimodal understanding. For developers, this model is therefore relevant to code assistance, autonomous or semi-autonomous agents, tool-integrated applications, and tasks that combine text with images. The model record lists image input and tool-calling as supported, while its use-case metadata also includes audio; the supplied material does not provide further details about audio handling, supported formats, output modalities, or tool-calling interface requirements. No context-window value is available in the supplied material, so capacity for long prompts and documents cannot be assessed here. Inference speed, pricing, and deployment requirements are also unavailable. The source excerpt identifies the Ollama distribution page but does not provide licensing terms, so review the applicable terms before redistribution or commercial deployment. This record does not include benchmark results or task-specific accuracy measurements, meaning the advertised positioning should not be treated as a substitute for evaluation on your own coding, vision, reasoning, or agent workloads.
- Model size
- 26B parameters
- Context
- Not listed
beam
beam is an open model by Reflection aimed at developers building coding workflows, agentic systems, and general reasoning applications. Its advertised strengths include agentic tool use and token-efficient usage, and the supplied record lists tool-calling as a capability. This makes it a candidate for assistants that need to invoke tools, coordinate multi-step tasks, generate or review code, or support reasoning-oriented workflows. The source does not list any supported modalities, so the supplied material provides no basis for treating beam as a vision, audio, image, or video model. Its context window, parameter count, inference speed, and pricing are not available in the supplied material, so deployment sizing and performance planning require separate validation. The source excerpt also does not provide licensing terms, safety classifications, benchmark results, or runtime configuration guidance; review applicable model documentation before production use. Overall, beam appears most relevant when tool-calling, coding, agents, agentic behavior, and reasoning are central requirements, but developers should test its actual reliability, latency, resource needs, and compatibility with their tool interfaces in the intended environment.
- Model size
- Size not listed
- Context
- Not listed
glm-5.3-flash
glm-5.3-flash is a multimodal model aimed at developers building coding tools and agentic workflows. The supplied material describes it as Z.ai’s first natively multimodal model and says it approaches Claude Opus 4.8 on coding and agentic benchmarks. It is listed with 18 billion active parameters, supports image input, and has tool-calling capability, making it relevant for code generation, visual understanding, and applications where the model can invoke external tools. Its advertised practical focus is coding, agents, agentic tasks, and vision. The supplied material does not provide a context-window size, inference-speed result, input or output pricing, or licensing terms, so those should be verified before deployment. It is presented through the Ollama model library, which may suit developers who want to run or integrate it through Ollama, but the excerpt does not specify hardware requirements, quantization options, deployment limits, or production support. The benchmark comparison is an advertised positioning statement rather than a full evaluation, and no detailed benchmark scores or methodology are included.
- Model size
- 18B parameters
- Context
- Not listed
minimax-m3.1-flash
minimax-m3.1-flash is a practical Ollama option for coding, agentic workflows, tool-calling, vision-related tasks, and cybersecurity use cases. The supplied model record identifies image support and tool-calling capability, while the source describes improvements for agentic tasks, coding, Computer Use, and cybersecurity. It advertises online inference that is nearly three times faster and usage costs reduced by 70%, although no exact tokens-per-second benchmark or pricing figures are provided in the supplied material. The model may be worth evaluating for developers building coding assistants, autonomous agents, computer-use workflows, or image-aware applications, but the available information does not establish how those capabilities perform in a particular runtime or hardware environment. Context-window size and parameter count are not available in the supplied material, so memory requirements and long-context suitability cannot be assessed from this record. Licensing terms are also not available in the supplied material and should be verified before commercial deployment. The source excerpt provides a relative speed and cost claim rather than reproducible benchmark details, so validate latency, quality, tool behavior, and vision performance against your workload.
- Model size
- Size not listed
- Context
- Not listed
north-mini-code-1.0
north-mini-code-1.0 is a coding-focused model intended for agentic software engineering and developer workflows. Ollama describes it as Cohere’s first model for developers and identifies it as a 30B Mixture-of-Experts model with 3B active parameters. Its advertised focus makes it relevant for writing, modifying, and navigating software projects, as well as building agents that can perform multi-step development tasks. Tool-calling is listed as a capability, so it may fit applications where an agent needs to invoke external tools during a coding workflow. The supplied material does not list supported modalities beyond its developer and coding focus. No context-window size, inference-speed benchmark, or pricing information is available in the supplied material, so those should be validated before deployment. Licensing terms are also not available in the supplied material and should be checked at the source before commercial or redistributed use. The model is available through Ollama at the supplied source URL, but runtime requirements and configuration details are not provided here.
- Model size
- 30B parameters
- Context
- Not listed
glm-5.3
glm-5.3 is positioned as Z.ai’s flagship open-weights model for coding and long-horizon agentic work. The supplied material describes it as the most capable open-weights model for coding and highlights major gains on long-horizon agentic tasks, making it a candidate for software development workflows, coding assistance, autonomous agents, and tasks that require extended multi-step execution. Tool-calling is explicitly listed as a capability, so it may be useful in applications where a model needs to invoke external tools as part of an agent workflow. The supplied material does not identify supported modalities beyond these text-oriented use cases, nor does it provide a context-window size, parameter count, inference-speed benchmark, or pricing information. It also does not include licensing terms, so review the applicable model and Ollama distribution terms before deploying it commercially or redistributing it. Because the description emphasizes coding and agentic performance rather than general-purpose feature coverage, evaluate it against your own requirements for reliability, tool integration, latency, and safety before production use.
- Model size
- Size not listed
- Context
- Not listed
qwen3.8:27b
qwen3.8:27b is a 27-billion-parameter model listed by Ollama for coding, agentic work, vision, and research. Its supplied capability information includes tool-calling, making it potentially useful for applications that invoke external tools or coordinate longer-running workflows. The source summary also highlights coding, professional work, research, and long-horizon agentic tasks, although no benchmark results or inference-speed measurements are provided in the supplied material. The model record identifies image modality support, so it may be relevant for image-aware developer workflows; the excerpt does not specify supported image formats, resolution limits, or whether image handling applies to all interfaces. An advertised context-window size is not available in the supplied material, and no runtime context configuration is provided. Pricing information is also unavailable. The supplied excerpt does not include licensing terms, deployment requirements, hardware guidance, safety classifications, or details about training data, so these should be verified before production use. For evaluation, focus on tool-calling reliability, coding quality, vision tasks, and behavior on long-horizon agent workflows relevant to your application.
- Model size
- 27B parameters
- Context
- Not listed
nemotron-cascade-2:30b
nemotron-cascade-2:30b is an open 30B mixture-of-experts model from NVIDIA aimed at reasoning and agentic workloads. The supplied description states that it has 30B total parameters and 3B activated parameters, so its sparse architecture may be relevant when evaluating deployment efficiency, although no hardware requirements or runtime performance figures are available in the supplied material. Its listed capability is tool-calling, making it a candidate for agents that need to invoke external tools or services as part of multi-step tasks. Reasoning, agents, and agentic workflows are the clearest practical use cases supported by the available information. The supplied material does not identify supported modalities, so modality-specific suitability cannot be confirmed. It also does not provide an advertised context-window size, inference speed benchmark, pricing, or licensing terms. Developers should therefore verify the model’s license and operational requirements before integrating it into a product, and should benchmark tool use and reasoning performance in their own environment. The source identifies the model as open, but that description alone does not establish a particular license or guarantee unrestricted use.
- Model size
- 30B parameters
- Context
- Not listed
qwen3.6:27b
qwen3.6:27b is a 27-billion-parameter model aimed at developers who need coding, agentic workflows, tool-calling, and image-oriented use cases. Its listed capabilities include tool-calling, while its recorded modality is image; the supplied model record also identifies coding, agents, agentic, and vision as intended use cases. The source description advertises substantial upgrades in agentic coding and preservation of thinking compared with previous Qwen models, but the supplied material does not provide independent benchmarks or an inference-speed measurement. A context-window value is not available in the supplied material, so deployment limits should be verified before configuring long prompts or extended agent sessions. Pricing information is also unavailable, and no licensing terms or license identifier are provided in the supplied material; review the model’s applicable terms before redistribution or commercial deployment. The Ollama page identifies this exact variant as qwen3.6:27b and shows an explicit update date of October 30, 2026. This makes it a plausible local Ollama option for coding assistants and tool-using agents, with image input or image-related workflows indicated, but performance, compatibility, and resource requirements should be tested in the target runtime.
- Model size
- 27B parameters
- Context
- Not listed