Thinking Machines Releases Weights for Inkling 2 Small
Thinking Machines Releases Weights for Inkling 2 Small
Inkling 2 Small is a 594B-parameter MoE model with 17B active parameters, trained with 12T tokens and released as open weights.
Thinking Machines Releases Weights for Inkling 2 Small
Inkling 2 Small is a mixture-of-experts (MoE) model with 594 billion total parameters and 17 billion active parameters. According to the release, pretraining was completed in just 68 hours using 12 trillion tokens. Reinforcement learning took an additional 9 days on a cluster of 2,048 GPUs.
The model is described as competitive with Chinese models from two to three months ago. Running it requires either two B300 GPUs or four H200 GPUs. A larger version is expected.
Why it matters
AnalysisThe release shows how quickly large MoE models can move from training to open weights, while also highlighting that serving such models still requires high-end GPU infrastructure.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Mistral AI (European Open and Efficient Models)
A series of high-performance open and commercial models from the French startup Mistral AI (Mistral 7B, Mixtral 8x7B, Mistral Large). Known for their compactness, speed, and respect for data privacy.
Open Weights vs Closed APIs
A comparison of two fundamental philosophies in AI distribution: Open Weights (Llama, Mistral, DeepSeek), which can be downloaded and run on your own server, versus Closed APIs (OpenAI, Anthropic, Google), where access is granted solely through subscription.
Tencent releases Hy4 Flash, a Claude Haiku competitor with 1M-token context
Tencent has released Hy4 Flash, a native multimodal model with a 1M-token context window and low inference costs. The company says it is comparable to Claude Haiku 5.5 in High mode for everyday tasks.
Part I. AI Agent Architecture: The Foundations
The eleven foundational building blocks of modern autonomous agents: model taxonomy, the agentic runtime loop, advanced prompting, distillation, skills architecture, memory systems, function calling, and FinOps.