# Thinking Machines Lab drops Inkling & Meta’s Muse Spark 1.1

## Executive summary

The AI landscape is shifting its focus from pure benchmark performance to customizable intelligence and architectural efficiency. This analysis covers three major model releases: Thinking Machines' open-weight Inkling (emphasizing customization via fine-tuning), Meta’s Muse Spark 1.1 (targeting cost-efficient agent workloads for enterprise use), and OpenAI's GPT-5.6 Sol, which showed progress on the challenging ARC-AGI-3 benchmark. Furthermore, Anthropic's J-space paper introduces a novel method to view internal model processing, offering potential new avenues for AI safety and control in agent development.

## Key takeaways

- Shift from Benchmarks to Customization: The industry debate is moving away from which closed model is the 'best' toward utilizing open base models combined with robust fine-tuning platforms (e.g., Inkling/Tinker API) for tailored, customizable intelligence.

## Technical details

- Thinking Machines Inkling Model Architecture: Inkling is an open-weight Mixture of Experts (MoE) model with a total parameter count of 975B and 41B active parameters. It features true multimodality, processing pixels down to the token level, and utilizes architectural choices for speed and efficiency.

## Practical implications

- **Agent Development & Control:** Anthropic's J-space research suggests a method to view internal model representations (the 'J-space'), which could provide a critical, localized, and editable internal state for monitoring agents before they execute actions. This is key for safety and control.
- **Cost-Effective Scaling:** Meta’s Muse Spark 1.1 emphasizes cost efficiency and scalability, making it attractive for running large-scale agent workloads in enterprise B2B SaaS environments.
- **Customization over Size:** The trend favors open-weight models that allow deep customization (via fine-tuning platforms) over simply adopting the largest frontier model available.

## Topics

Generative AI, Large Language Models (LLMs), Mixture of Experts (MoE), AI Safety, Agent Orchestration, Mixture of Experts podcast page, IBM AI Newsletter Signup

Source: https://www.youtube.com/watch?v=8rGYGFmytQs