Moondream: Moondream 3 Preview
moondream/moondream3-previewA preview of Moondream 3 with a compact MoE architecture; in Ailoot API, it responds to text questions and instructions.
A preview of Moondream 3
Moondream 3 Preview belongs to the Moondream family, developed around image understanding and tasks involving specific objects in a scene. For the third generation, the developer adopted a mixture-of-experts architecture: the model has 9 billion total parameters and uses approximately 2 billion per token. This combines a larger total set of weights with a smaller active portion when processing each token.
The official Preview model card describes 24 layers, a SigLIP-based vision encoder and the SuperBPE tokenizer. The underlying model’s specialized skills include image question answering, captioning, pointing and drawing bounding boxes around objects. These are distinct tasks: explaining scene content, describing an entire image and locating a specific object produce different kinds of results.
Text work in Ailoot API
This model is available in Ailoot API with text input and text output. Images, object coordinates and specialized visual methods are outside the listing’s current interface. The developer nevertheless provides a separate Preview example of a question without an image, in which the model explains a concept in response to an ordinary text instruction.
Use this mode for explanations, answers based on supplied information and restructuring text into a specified format. For example, provide a draft description, state the audience and request a clearer rewrite that preserves the facts. Alternatively, supply written instructions and ask for a sequence of actions with a short explanation of each step.
How to phrase a request
Name the task and response format explicitly: a brief explanation, connected prose, a list or a table. If you need analysis of specific material, include it in the message and identify the information that must remain unchanged. A question about a visual scene in text mode requires a verbal description; the model works with the supplied description rather than the image itself.
Moondream 3 Preview has a stated context of 32,000 tokens in Ailoot API, including messages and material in the current discussion. The Preview designation identifies it as a preliminary third-generation release; consider this version’s properties separately from version 3.1, especially when working with the underlying models’ specialized capabilities.
Available through Ailoot API
POST /v1/messages- Input types
text- Output types
text- Context
- Not specified in the catalogue
- Max. output
- Not specified in the catalogue
- Input
- $0.0006 / 1K
- Output
- $0.005 / 1K
POST /v1/responses- Input types
text- Output types
text- Context
- Not specified in the catalogue
- Max. output
- Not specified in the catalogue
- Input
- $0.0006 / 1K
- Output
- $0.005 / 1K
POST /v1/chat/completions- Input types
text- Output types
text- Context
- Not specified in the catalogue
- Max. output
- Not specified in the catalogue
- Input
- $0.0006 / 1K
- Output
- $0.005 / 1K
POST /v1beta/models/moondream/moondream3-preview:generateContent- Input types
text- Output types
text- Context
- Not specified in the catalogue
- Max. output
- Not specified in the catalogue
- Input
- $0.0006 / 1K
- Output
- $0.005 / 1K
moondream/moondream3-preview · JSON
{"about":"### A preview of Moondream 3\n\nMoondream 3 Preview belongs to the Moondream family, developed around image understanding and tasks involving specific objects in a scene. For the third generation, the developer adopted a mixture-of-experts architecture: the model has 9 billion total parameters and uses approximately 2 billion per token. This combines a larger total set of weights with a smaller active portion when processing each token.\n\nThe official Preview model card describes 24 layers, a SigLIP-based vision encoder and the SuperBPE tokenizer. The underlying model’s specialized skills include image question answering, captioning, pointing and drawing bounding boxes around objects. These are distinct tasks: explaining scene content, describing an entire image and locating a specific object produce different kinds of results.\n\n### Text work in Ailoot API\n\nThis model is available in Ailoot API with text input and text output. Images, object coordinates and specialized visual methods are outside the listing’s current interface. The developer nevertheless provides a separate Preview example of a question without an image, in which the model explains a concept in response to an ordinary text instruction.\n\nUse this mode for explanations, answers based on supplied information and restructuring text into a specified format. For example, provide a draft description, state the audience and request a clearer rewrite that preserves the facts. Alternatively, supply written instructions and ask for a sequence of actions with a short explanation of each step.\n\n### How to phrase a request\n\nName the task and response format explicitly: a brief explanation, connected prose, a list or a table. If you need analysis of specific material, include it in the message and identify the information that must remain unchanged. A question about a visual scene in text mode requires a verbal description; the model works with the supplied description rather than the image itself.\n\nMoondream 3 Preview has a stated context of 32,000 tokens in Ailoot API, including messages and material in the current discussion. The Preview designation identifies it as a preliminary third-generation release; consider this version’s properties separately from version 3.1, especially when working with the underlying models’ specialized capabilities.","alias_of":null,"aliases":[],"author":"Moondream","canonical_model":"moondream/moondream3-preview","ctx":null,"formats":["chat","gemini","messages","responses"],"id":"moondream/moondream3-preview","input":["text"],"kind":"text","max_out":null,"methods":[{"cache_read":null,"cache_write":null,"cache_write_1h":null,"cache_write_5m":null,"conditional_meter":null,"ctx":null,"family":"anthropic","fast_multiplier":1,"flex_multiplier":1,"fmt":"messages","http":"POST","img_in":null,"img_out":null,"in":0.6,"input":["text"],"label":"Anthropic Messages","max_out":null,"max_tool_calls":null,"out":5.0,"output":["text"],"path":"/v1/messages","request_model_id":"moondream/moondream3-preview","route_label":"Ultra","status_reads":[],"tiers":[],"tool_prices":null,"unit":"token","units":[]},{"cache_read":null,"cache_write":null,"cache_write_1h":null,"cache_write_5m":null,"conditional_meter":null,"ctx":null,"family":"openai","fast_multiplier":1,"flex_multiplier":1,"fmt":"responses","http":"POST","img_in":null,"img_out":null,"in":0.6,"input":["text"],"label":"OpenAI Responses","max_out":null,"max_tool_calls":null,"out":5.0,"output":["text"],"path":"/v1/responses","request_model_id":"moondream/moondream3-preview","route_label":"Ultra","status_reads":[],"tiers":[],"tool_prices":null,"unit":"token","units":[]},{"cache_read":null,"cache_write":null,"cache_write_1h":null,"cache_write_5m":null,"conditional_meter":null,"ctx":null,"family":"openai","fast_multiplier":1,"flex_multiplier":1,"fmt":"chat","http":"POST","img_in":null,"img_out":null,"in":0.6,"input":["text"],"label":"OpenAI Chat","max_out":null,"max_tool_calls":null,"out":5.0,"output":["text"],"path":"/v1/chat/completions","request_model_id":"moondream/moondream3-preview","route_label":"Ultra","status_reads":[],"tiers":[],"tool_prices":null,"unit":"token","units":[]},{"cache_read":null,"cache_write":null,"cache_write_1h":null,"cache_write_5m":null,"conditional_meter":null,"ctx":null,"family":"gemini","fast_multiplier":1,"flex_multiplier":1,"fmt":"gemini","http":"POST","img_in":null,"img_out":null,"in":0.6,"input":["text"],"label":"Gemini generateContent","max_out":null,"max_tool_calls":null,"out":5.0,"output":["text"],"path":"/v1beta/models/moondream/moondream3-preview:generateContent","request_model_id":"moondream/moondream3-preview","route_label":"Ultra","status_reads":[],"tiers":[],"tool_prices":null,"unit":"token","units":[]}],"model_aliases":[{"canonical_model_id":"moondream/moondream3-preview","request_model_id":"moondream/moondream3-preview","route_label":"Ultra"}],"name":"Moondream: Moondream 3 Preview","price":{"in":0.6,"in_label":null,"out":5.0,"out_label":null,"sort_in":0.6,"sort_out":5.0,"varies":false},"public_id":"moondream/moondream3-preview","short":"A preview of Moondream 3 with a compact MoE architecture; in Ailoot API, it responds to text questions and instructions."}