Featherless Serverless LLM Inference

Featherless AI is a Singapore-based provider of serverless inference for open-source AI models, according to press coverage dated May 2026. It serves developers and companies that want to run many open models without managing GPU infrastructure, and its team contributes to the RWKV project.

The service gives access to 40,000+ open models through one API key. Plans range from a flat monthly chat plan to a developer plan with per-token credits and a business plan on annual contracts with dedicated GPUs (H100, MI325, B200, B300), burst and failover to public cloud, and fine-tuning support. A tool called Jev turns any model into a classifier.

• Model catalogue: 40,000+ open models behind a single API key.
• Dedicated GPUs: H100, MI325, B200 and B300 under custom business plans.
• Burst and failover: overflow to public cloud for reliability.
• Jev classifier tool: converts any model into a classifier.
• Fine-tuning support: customised models to reduce cost over time.
• Flexible plans: chat, developer (per-token credits) and business tiers.

Vietnamese software companies can prototype chatbots and text classification on different open models via one API. Startups can build classifiers for support tickets or content moderation using Jev. Enterprises can test open models before buying dedicated GPUs.

One API key reaches a very large catalogue of open models, avoiding separate hosting per model. Business plans offer dedicated GPUs with failover, and fine-tuning that can lower cost over time. Per-token developer billing suits experiments before committing to annual contracts.

Companies are shifting from single proprietary models to mixed open-model stacks, creating demand for managed inference. Cost and data-control concerns push Southeast Asian developers toward open models. Vietnamese AI teams building products on open LLMs can use hosted inference without GPU investment.

Singapore-based per KrASIA (May 2026); the website does not state an address. USD 20 million Series A co-led by AMD Ventures and Airbus Ventures. Team contributes to RWKV, a Linux Foundation project.