Moondream 3 Preview and Moondream 3.1 from @moondreamai are now on OpenRouter, served by @PrimeIntellect
Object detection and pointing work in plain chat: attach one image and start your message with /detect {"object":"coffee cup"} or /point {"object":"croissant"}
Object detection and pointing work in plain chat: attach one image and start your message with /detect {"object":"coffee cup"} or /point {"object":"croissant"}
Oct 15, 2026 · 11:23 PM UTC
10
8
82
21,809
Small vision-language models (9B total, 2B active MoE): image + text in, 32K context, optional reasoning. Use /caption {"length":"normal"} for captions
Both are $0.30/M input and $2.50/M output tokens
See more:
openrouter.ai/moondream/moon…
Both are $0.30/M input and $2.50/M output tokens
See more:
openrouter.ai/moondream/moon…
1
0
8
3,879
@OpenRouter @moondreamai @PrimeIntellect pointing at the thing beats describing where it is. so much easier to check
0
0
1
13
@OpenRouter @moondreamai @PrimeIntellect this is 2.5x cheaper on Cloudflare developers.cloudflare.com/workers-ai/mod…
0
0
1
91
@OpenRouter @moondreamai @PrimeIntellect object detection in plain chat is the feature nobody asked for until they see it. moondream turning 'point at the thing' into an api call is quietly huge for agents
0
0
0
7
@OpenRouter @moondreamai @PrimeIntellect "Works in plain chat" is doing a lot of work in that sentence. A slash command with a JSON payload is a CLI in a trenchcoat. Nothing wrong with a CLI — just don't call it plain.
0
0
0
123
@OpenRouter @moondreamai @PrimeIntellect Plain chat syntax for bounding boxes and pointing without parsing raw coordinate JSON is so clean. Definitely gonna try this for quick visual automation scripts!
0
0
0
61
