Nebius Token Factory
Features
- Runs 60+ open-source models such as DeepSeek V3, Llama 3.3 and Qwen 3
- Fast and Base serving options for the same model
- Dedicated GPU endpoints with 99.9% uptime SLAs
- Fine-tuning workflow with LoRA or full-tuning and private deployment
- Per-token pricing instead of GPU-hour billing
- Text, code and vision models through a unified API
- Team management, SSO and fine-grained access control
- Speculative decoding for high throughput
Pricing Details
💵 Pricing (January 2026 Status) Nebius Token Factory offers some of the most competitive rates in the market, often 2.5x cheaper than proprietary alternatives like GPT-4o for comparable quality: Model Category Input Price (per 1M) Output Price (per 1M) Small Models (e.g., Llama 8B) ~$0.02 ~$0.06 Mid-Range (e.g., Llama 70B) ~$0.13 ~$0.40 Frontier (e.g., DeepSeek V3) ~$0.50 ~$1.50 Batch Inference 50% Discount 50% Discount 🎁 Free Version: Nebius offers $1 in free credits to all new users, which allows you to process millions of tokens on smaller models in their "Playground" or via API without any initial commitment.