NewsBite analysis · AI, Data, Programming
The Decoder
Thinking Machines bets on efficiency over size with its second model, Inkling Small
Thinking Machines, led by former OpenAI CTO Mira Murati, has unveiled a new AI model named Inkling Small. Despite having fewer parameters than its predecessor, it achieves impressive scores in various coding and reasoning tests, successfully balancing performance and efficiency. This approach signals a shift in AI model development towards optimizing size and resource usage while ensuring high performance in specific tasks. Such advancements suggest that AI might be moving towards models that provide more value for less computational cost.
MonitorConfidence: HighShort term horizon
Published 31 Jul 2026, 18:10 UTC1 min read
Visit original article ↗Lead image — The Decoder
Summary
TL;DR
- Inkling Small scores 40 on the Intelligence Index, close to its predecessor Inkling.
- Achieves greater token efficiency, averaging 24K output tokens per task.
- Competes well in specific reasoning tasks compared to larger models.
- Weights available for fine-tuning on Hugging Face.
- Model aims to be a foundation for personalized data adaptation.
Interesting facts
- Inkling Small has 276 billion total parameters but only 12 billion active ones.
- It surpasses larger models in coding efficiency and testing performance.
- Uses a 256K-token context window, showcasing advanced processing capabilities.