FAQ
Questions about Varion.
Access, trials, commercial adoption and validation.
Can I access each Varion product separately?
Yes. Forge, Compute and Token Optimisation each have a dedicated sign-in entry and can be adopted independently.
Can I test before buying?
Yes. Forge and Token Optimisation include free trial access, and Compute includes a free self-test path for commercial evaluation.
What has Varion Compute validated so far?
Current completed NVIDIA A40 tests include real LLM workloads up to Qwen 2.5 1.5B, with up to 36.4% lower energy per generated token, up to 131.4% higher generation throughput and 100% identical greedy-generation output in the cited KV-cache tests.
Are Compute results guaranteed on every GPU?
No. Results vary by model, workload and hardware. The current cited evidence is from NVIDIA A40 validation; broader cross-GPU validation is ongoing.
How do I discuss a commercial pilot?
Use the contact page for a scoped commercial evaluation.