Can I run Inkling locally?
Inkling by thinkingmachines has 952B parameters (19B active per token, 256 experts, 6 used per token). At 4-bit with an 8K context it needs about 545 GB of GPU memory. It fits on 1 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/inkling-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | rough | 334 GB | 335 GB | 342 GB | 366 GB |
| 3-bit (Q3_K_M) | usable | 435 GB | 436 GB | 443 GB | 467 GB |
| 4-bit (Q4_K_M) | good, most popular | 544 GB | 545 GB | 551 GB | 576 GB |
| 5-bit (Q5_K_M) | very good | 634 GB | 635 GB | 641 GB | 666 GB |
| 6-bit (Q6_K) | near lossless | 729 GB | 730 GB | 736 GB | 761 GB |
| 8-bit (Q8_0) • | effectively lossless | 800 GB | 801 GB | 807 GB | 832 GB |
| 16-bit (F16/BF16) • | original weights | 1766 GB | 1767 GB | 1773 GB | 1798 GB |
Which GPUs and Macs run Inkling?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Prefer an API?
Thinking Machines: Inkling is on OpenRouter at $1.00 per 1M input tokens and $4.05 per 1M output tokens. Price details and history →
Questions
How much VRAM does Inkling need?
About 545 GB at 4-bit (Q4_K_M) with an 8K context: 542 GB of weights (estimated), 2.1 GB of KV cache and 0.8 GB of runtime overhead. At 8-bit it needs about 801 GB.
What is the smallest GPU that runs Inkling?
No single consumer GPU in our table fits it at Q4_K_M with 8K context.
Can I run Inkling on a Mac?
Not at Q4_K_M with 8K context on any Mac in our table, even a 512 GB Mac Studio, using the default GPU memory limit.
Model facts
| Hugging Face repo | thinkingmachines/Inkling |
| Parameters | 952,377,623,626 (952B) · ~19B active (computed from config), low confidence, speed not shown |
| Layers / hidden size | 66 / 6144 |
| Attention | GQA; 64 heads, 8 KV heads, head dim 128 |
| KV cache per token (FP16) | 264.0 KiB |
| Max context | — |
| License | apache-2.0 |
| Released on HF | Jul 14, 2026 |
| Downloads (30d) / likes | 1,451,864 / 1,853 |
Sources: Hugging Face model API · config.json (thinkingmachines/Inkling) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.