Inkling mixture-of-experts model enables token-efficient reasoning

Summary

Developers can now dial reasoning depth per-request without infrastructure management, and handle text+image+audio in one model call instead of chaining separate services. The controllable inference effort directly reduces token spend and latency costs at request granularity.

Why it matters

Developers can now dial reasoning depth per-request without infrastructure management, and handle text+image+audio in one model call instead of chaining separate services. The controllable inference effort directly reduces token spend and latency costs at request granularity.

Sources

1. 1.975B total parameters, 40B active parameters per token, and a context window of 1M tokens
2. 2.Inkling accepts text, image, and audio inputs and produces text outputs through a unified decoder architecture
3. 3.Developers can adjust inference effort to balance reasoning depth, token usage, and latency for different workloads
4. 4.Inkling is live on Together AI Serverless today. No waiting for capacity, no infrastructure to provision, no GPUs to manage
5. 5.mixture-of-experts architecture with a shared expert sink