ApX logoApX logo

Moondream 3.1

Active Parameters

9B

Context Length

32K

Modality

Multimodal

Architecture

Mixture of Experts (MoE)

License

other

Release Date

30 Jun 2026

Knowledge Cutoff

-

API Pricing (per 1M)

Input: $0.30 · Output: $2.50

System Requirements

VRAM requirements for different quantization methods and context sizes

1,024 tokens

19.23 GB VRAM

Consumer

1x RTX 4090

24GB VRAM

Datacenter

1x NVIDIA A100

80GB VRAM

Apple Silicon

1x Apple M3 Max

128GB VRAM

32,000 tokens

23.20 GB VRAM

Consumer

1x RTX 4090

24GB VRAM

Datacenter

1x NVIDIA A100

80GB VRAM

Apple Silicon

1x Apple M3 Max

128GB VRAM

Architecture Diagram

Input TokensToken EmbeddingContext: 32Kx N layersNormPre-AttentionSelf-Attention+NormPre-FFNSparse MoE FFNActivation+Final NormOutput Logits

Evaluation Benchmarks

No evaluation benchmarks for Moondream 3.1 available.

Rankings

Overall Rank

-

Coding Rank

-

About Moondream 3.1

Moondream 3.1 is an open-weight mixture-of-experts vision-language model with 9B total and 2B active parameters per token. It delivers efficient visual question answering, captioning, optical character recognition, and visual reasoning in constrained environments.

Technical Specifications

Mixture of Experts

Total Expert Parameters

2.0B

About Moondream

The Moondream model family developed by Moondream.


Other Moondream Models
  • No related models available