Thinking Machines Lab · Looped MoE Multimodal
Inkling 2 Mini
Inkling 2 Mini is an open-weight multimodal model that accepts text and image inputs and produces text, released under Apache 2.0 with a 1M-token context window. It has 26.4B total parameters (3.4B active) and uses a looped transformer design in which the middle 12 layers run three times, giving about 5.4B effective active parameters per token. The lab reports 79.2% on SWE-bench Verified and 71.6% on Terminal Bench v2.1, with the checkpoint (about 20GB, NVFP4 and MXFP8) runnable on a DGX Spark or 32GB-plus Apple silicon via vLLM, SGLang, llama.cpp and Ollama.
Open the source ↗Similar Language Models releases
- ERNIE 5.5 Flash Preview
Baidu · Agentic MoE coding model · Week 47, 2026
- Edge0-145B-NVFP4
Edge0 (Edge0-AI) · Latent MoE LLM · Week 47, 2026
- LongCat-2.5
Meituan (LongCat team) · Agentic MoE LLM · Week 47, 2026
- Inkling 2 Small
Thinking Machines Lab · MoE agentic LLM · Week 46, 2026
- Kimi K3.1 Pro
Moonshot AI · Open agentic MoE · Week 46, 2026
- Inkling 2 Small
Thinking Machines Lab · MoE Agentic Model · Week 46, 2026
More from Thinking Machines Lab
- Inkling 2 Small
Thinking Machines Lab · MoE agentic LLM · Week 46, 2026
- Inkling 2 Small
Thinking Machines Lab · MoE Agentic Model · Week 46, 2026
- Inkling-Small
Thinking Machines Lab · MoE Architecture · Week 31, 2026
- Interaction Models
Thinking Machines Lab · MoE Architecture · Week 30, 2026
- Inkling
Thinking Machines Lab · MoE Architecture · Week 29, 2026
Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.