thinkingmachines/Inkling-2-Small
just landedmodel
Inkling 2 Small is a general-purpose multimodal model that accepts text and image inputs and generates text outputs. Mixture-of-experts, 594B total, 17B active (384 experts, 8 per token). Context: 1M tokens (config). License: apache-2.0.
- Org
- thinkingmachines
- Released
- 2026-11-04
- Status
- just landed
- License
- apache-2.0 (permissive)
- Parameters
- 594.1B total, 17.0B active
- Architecture
- MoE · 384 experts, 8/token · Inkling2ForConditionalGeneration
- Context
- 1,048,576 tokens (config.json)
- Task
- image-text-to-text · vision
- Signal
- 330 HF likes · 7329 downloads
Runs on
mixed: real files + estimate; need = weights + FP16 KV cache for 8K tokens + 1GB. A sizing calculation, not tested.
- Q4 ~356.8 GB: Mac 512GB
- Q8 ~389.1 GB: no listed device
- FP16 ~1190.7 GB (estimated size): no listed device
Links
- Hugging Face
- Ollama
- MLX: mlx-community/Inkling-2-Small-4bit
- MLX: mlx-community/Inkling-2-Small-oQ4e
- GGUF: unsloth/Inkling-2-Small-GGUF
First published here 2026-11-04 · updated 2026-11-13 · markdown · JSON