by Thinking Machines

Inkling FP4.

text open weights datacenter 952B params 524K ctx
Cheapest input
$1.0/M
on Together AI
Cheapest output
$4.05/M
on Together AI
Smallest GPU
8× Nvidia A100 SXM4
$5.73/hr

Inkling is an Apache 2.0 mixture-of-experts model, 975B total and 41B active, that takes text, images and audio and returns text, with weights in BF16 and NVFP4.

Smallest GPU to run it See all quantisations →

8× Nvidia A100 SXM4 · $5.73/hr.

Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.

Where to use it

Cheapest hosted endpoints.

Provider Access $/M in $/M out
Together AI hosted inference $1.0 $4.05 Launch ↗
OpenRouter api aggregator $1.0 $4.05 Launch ↗
Sources

Official references.

FAQ

Frequently asked.

How do I run Inkling FP4?
Inkling FP4 is open-weight, so you can self-host on rented GPUs. See the Run It Yourself tab for GPU configurations + cost estimates, or use one of the hosted inference providers listed on this page.
Where can I access Inkling FP4?
Inkling FP4 is available via Fireworks AI, Together AI, OpenRouter. Each access option lists its own pricing (per million tokens or hourly hosting).
How much does it cost to run Inkling FP4?
API pricing starts at $1.0/M input tokens and $4.05/M output tokens. Self-hosting cost depends on the GPU you rent — see the Run It Yourself tab.
Is Inkling FP4 open-source or proprietary?
Inkling FP4 is open-weight under the apache-2.0 license. You can download and self-host it.