Thinking Machines’ Inkling, released July 15, 2026 under Apache 2.0, is a 975B MoE model with 41B active parameters. It directly challenges GPT-5.6 Sol.
Inkling
$1.87/$4.68 per million tokens. Apache 2.0 license. MoE architecture for efficiency. Any organization can inspect, modify, and deploy.
GPT-5.6 Sol
$5/$30 per million tokens. Proprietary. Mature API ecosystem, function calling, structured outputs.
Significance
Inkling proves open-source competes at the frontier. The Apache 2.0 license has profound implications for AI sovereignty and enterprise adoption. For organizations wanting control and transparency, Inkling is a compelling alternative to proprietary flagships.
Related: AI Model Comparison 2026 · Best Free AI Models
