Inkling FP4.
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.
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.
Cheapest hosted endpoints.
Variants in the Inkling family.
Cheapest hardware per quantisation.
Each row is one quantisation tier (the same weights compressed differently). Lower precision → lower VRAM → cheaper hardware, at the cost of small accuracy loss. $/hr is the cheapest rate on offer now. A multi-card setup uses a real instance of that size when one is rented; otherwise it is the single-card rate times the count.
| Quantisation | Cheapest GPU config | Total VRAM | Live $/hr | tokens/sec | |
|---|---|---|---|---|---|
|
FP16
FP16 — half precision (default)
|
2304 GB |
$68.80/hr
Oracle Cloud · on-demand
|
— | Compare → | |
|
FP8
FP8 — 8-bit float (Hopper / Blackwell)
|
1536 GB |
$22.54/hr
Cyfuture AI · on-demand
|
— | Compare → | |
|
INT4
INT4 — 4-bit integer (~4× VRAM saving)
|
640 GB |
$5.73/hr
est. 8 × single-card rate
Nosana · on-demand
|
— | Compare → |
What it costs per month across providers.
Estimate your monthly bill for Inkling FP4 across every host that publishes per-token pricing. Slide your token volumes; the chart + table re-rank cheapest-first.
Cheapest provider on the left.
Total monthly cost — input + output tokens combined.
Bill breakdown.
Rent the GPU instead of paying per token.
For an open-weights model like Inkling FP4, you can rent a GPU and serve inference yourself. The math: cheapest GPU rental × 730 hours/month. Power is included in a rental's hourly price.
Assumes the GPU runs 24/7 at ~85% utilisation. If your traffic is bursty, you'll pay less for the API and probably more for the GPU (idle hours still cost rental). The breakeven analysis lives on the Self-host vs API breakeven tool.
About Inkling FP4.
Inkling is an Apache 2.0 open-weights model from Thinking Machines that takes text, images and audio and returns text. It has 975 billion parameters, 41 billion active, across 66 layers, routing each token to 6 of 256 experts plus 2 shared ones. Audio should be 16kHz WAV, ideally under 20 minutes. Weights come in BF16 and NVFP4. The maker aims it at developers building agents, coding assistants, chatbots and retrieval systems. It warns of hallucination and weaker results in long conversations, and advises human review for high-stakes use.