Moondream 3 Preview
Summary
Moondream 3 Preview is an open-weight language model from Moondream, released on 15 Oct 2026. It accepts text and images and generates text, with a 32K-token context window and up to 4K output tokens per response. As of 17 Oct 2026, the lowest listed API price is $0.30 per million input tokens and $2.50 per million output tokens (OpenRouter), across 1 provider we track.
Moondream 3 Preview is a vision-language model from Moondream with 9B total parameters and 2B active (mixture-of-experts). It handles visual question answering, captioning, OCR, and visual reasoning over images, with...
Last verified 17 Oct 2026. Every figure on this page links to its source.
Specifications
| Developer | Moondream |
|---|---|
| Release date | 15 Oct 2026 [OpenRouter] |
| Status | Generally available |
| Weights | Open weights [Hugging Face] |
| Parameters | 9.3B [Hugging Face] |
| Context window | 32K (32,000 tokens) [OpenRouter] |
| Max output | 4K tokens [OpenRouter] |
| Input | Text and images [OpenRouter] |
| Output | Text [OpenRouter] |
| Reasoning mode | Yes [OpenRouter] |
| Tool calling | No [OpenRouter] |
| Structured output | No [OpenRouter] |
| Hugging Face | moondream/moondream3-preview [Hugging Face] |
Moondream 3 Preview API pricing by provider
USD per million tokens. Sorted by input price.
| Provider | Input | Output | Cached input | Context | Source |
|---|---|---|---|---|---|
| OpenRouter moondream/moondream3-preview | $0.30 | $2.50 | – | 32K | OpenRouter, 17 Oct 2026 |
Running Moondream 3 Preview locally
Estimated memory, from the parameter count on Hugging Face. Includes about 0.8 GB of runtime overhead.
| Quantization | Weights | Total, 8K context | Total, 32K context |
|---|---|---|---|
| Q4_K_M (4-bit) | 5.6 GB | 6.4 GB | – |
| Q8_0 (8-bit) | 9.8 GB | 11 GB | – |
| FP16 / BF16 (16-bit) | 19 GB | 19 GB | – |
At 4-bit with an 8K conversation, Moondream 3 Preview needs about 6.4 GB, so it fits on a graphics card with 8 GB or more, such as the NVIDIA GeForce RTX 4060 8 GB. The model's architecture is not public, so the conversation cache is not included. Check your own hardware.
Frequently asked questions
What is the context window of Moondream 3 Preview?
Moondream 3 Preview has a context window of 32,000 tokens (32K) and can generate up to 4,096 tokens in a single response, according to OpenRouter.
How much does the Moondream 3 Preview API cost?
Moondream 3 Preview is listed from $0.30 per million input tokens and $2.50 per million output tokens (OpenRouter). Prices last verified 17 Oct 2026.
Is Moondream 3 Preview open source?
Moondream 3 Preview is an open-weight model: its weights are published on Hugging Face as moondream/moondream3-preview. Check the license terms before commercial use.
When was Moondream 3 Preview released?
Moondream released Moondream 3 Preview on 15 Oct 2026, according to OpenRouter.
How much VRAM does Moondream 3 Preview need?
About 6.4 GB at 4-bit (Q4_K_M) with an 8K-token conversation, and about 11 GB at 8-bit. The 4-bit model fits on a graphics card with 8 GB or more, such as the NVIDIA GeForce RTX 4060 8 GB. These are estimates from the model's published parameter count.
Change history
- Parameter count changed from none to 9.3B
- Added to the directory
- Now listed on OpenRouter