Can You Run Inkling Locally?
Author · AI Local Check · Published July 17, 2026
Inkling is the first model from Thinking Machines, the lab founded by former OpenAI CTO Mira Murati — and it landed on July 15, 2026 as an open-weight release. It's already on Hugging Face as GGUF, which raises the obvious question: can you run Inkling locally?
The honest answer: not on a normal PC — Inkling is enormous — but, unlike some headline announcements, its weights really are available to download today.
What is Inkling?
Inkling is a Mixture-of-Experts (MoE) model with close to 950 billion parameters (around 41 billion active per token), trained to reason natively across text, image, audio and video. It is open-weight, so anyone can download it — the real constraint is whether your hardware can hold it in memory.
• Maker: Thinking Machines (Mira Murati)
• Released: July 15, 2026 (open weights, on Hugging Face)
• Size: ~950B parameters (MoE, ~41B active)
• Multimodal: text, image, audio, video
How much memory does Inkling need?
Computed from its real GGUF files, even the most aggressive quantizations are far beyond a desktop:
| Quantization | Memory to load | Verdict |
|---|---|---|
| IQ1_M (~1-bit) | ~226 GB | Beyond any consumer GPU |
| ~4-bit | ~498 GB | Server-scale |
| Q8_0 (8-bit) | ~953 GB | Data-center |
Even at an extreme 1-bit quantization, Inkling needs about 226 GB — more than a stack of RTX 5090s. It's designed for multi-GPU servers or large unified-memory machines, not a gaming PC.
What can you run instead?
If you want a capable model on your own GPU today, plenty fit consumer cards. See the best LLM for your VRAM for size-by-size picks — models from the Qwen, Llama, Mistral and DeepSeek families run well locally. Inkling joins Kimi K3 and DeepSeek V4 in the "too big for home" tier of 2026's giant open models.
The bottom line
Inkling is a landmark open release from a major new lab — genuinely downloadable, but genuinely huge (~950B, ~226 GB even at 1-bit). Run it only if you have server-class hardware. For everything else, check what your PC can actually run and pick a model that fits.