GPT-OSS 120B
Dedicated GPUs for
open-source AI
Zero DevOps
Your per-token bill is scaling faster than your users?
Dedicated capacity is the fix. But GPUs without a team to run them is a hire you didn't budget for. We give you both.
Hardware
B300, B200, RTX Pro 6000, MI 325 RAM, Model sizes supported, token volumes supported, Price Contact us
Why Dedicated Infrastructure with Featherless AI
Available today
No waitlist. No procurement cycle.
Dedicated capacity ready to deploy now — no waitlist, no months-long procurement cycle. Provision in the US, EU, or Southeast Asia as your users and compliance require
AMD partner
Leading price-per-compute on AMD
As an AMD partner, we've already done the work of optimizing leading open models to run fast on AMD accelerators. That means you get to take advantage of leading price-per-compute on AMD hardware.
Expert optimization
Built by the team behind RWKV.
We tune the full inference stack: quantization, batching, fine-tuning, and distillation. So you can use your proprietary data to improve performance and lower costs.
Guaranteed performance
Benchmark first. Then lock the capacity.
Benchmark your workload with us, then lock that capacity for yourself alone. Compute entirely dedicated to you. Reserved, isolated, and never reallocated to someone else.
Privacy & compliance
Your data and models stay yours.
Your data and models stay yours, on dedicated, isolated infrastructure. VPC isolation for sensitive workloads, plus the compliance controls production teams require.
FAQs
How is dedicated GPU pricing calculated?
Dedicated GPU pricing is based on the hardware tier you reserve and the duration of your reservation. No per-token billing — you pay a fixed rate for exclusive access to your compute. Pricing varies by GPU model, quantity, and region. Contact us for a custom quote →
What does optimisation and MLOps support include?
Our team handles the full inference stack — quantisation, batching, fine-tuning on your real traffic, and ongoing distillation. On-call support and proactive tuning included. Not a separate engagement.
What performance guarantees and SLAs do you offer?
We benchmark your specific workload before you commit, then guarantee that performance level on reserved capacity. SLA terms are defined per contract. Talk to us →
How does Featherless handle data privacy and compliance?
Your data never leaves your dedicated environment. VPC-level isolation — prompts, completions, and model weights not shared with any other customer.
Where is the infrastructure hosted? Can I choose a region?
Available in the US, EU, and Southeast Asia. You choose the region that matches your users and data residency requirements. Multi-region deployments supported. Speak with our team →