Skip to main content
ScitiX Model Inference is a high-efficiency, reliable, and cost-effective model API platform for developers and enterprises. It brings multiple model categories behind a unified, OpenAI-compatible API, with inference acceleration, elastic scaling, built-in observability, and enterprise-grade security to help you move from prototyping to production.

​
Core Capabilities

  • Unified API: OpenAI-compatible REST and streaming interfaces for text generation, vision/multimodal, embeddings, and speech, so you can integrate with mainstream SDKs and coding tools with minimal changes.
  • High-performance inference: In-house inference acceleration and auto-scaling support low latency, high throughput, and stable long-context workloads.
  • Rich model catalog: Leading open-source models across multiple modalities, available through a single API and the Models page.
  • Cost control: Pay-as-you-go billing with elastic scaling, plus per-key quotas and TPM/RPM rate limits to manage spending.
  • Observability and security: View token consumption, spending, request details, and performance metrics from Home, Usage Detail, and Metrics; manage API keys across their lifecycle, including pause and spending caps.

​
Platform Entry Points

  • Model Selection & Playground: Get model recommendations, compare models on the Models page, then test prompts and parameters in the Playground.
  • First API Call: Create an API Key and call models through the API.
  • View Usage: Review request-level details, token usage, cost, and latency metrics. For aggregated metrics and trends, see View Metrics.

​
Available Models and Coverage

The platform currently offers models across text generation, reasoning, vision/multimodal, embeddings, and speech. For the latest model list and pricing, refer to the Models page in the console.
The model catalog is updated as newer versions are released.

​
Typical Use Cases

  • Intelligent Q&A and assistants: Build enterprise knowledge Q&A, operations assistants, and customer service bots with general-purpose LLMs such as Qwen/Qwen3-32B or DeepSeek-V4-Flash.
  • Content creation and ideation: Generate marketing copy, long-form content, and structured documents.
  • Code intelligence and developer productivity: Use models such as glm-5.2, Qwen/Qwen3.6-27B, and MiniMaxAI/MiniMax-M2.7 for agentic coding, refactoring, and multi-step development workflows.
  • RAG (Retrieval-Augmented Generation): Build retrieval pipelines with embedding models such as Qwen/Qwen3-Embedding-8B, Qwen/Qwen3-Embedding-4B, Qwen/Qwen3-Embedding-0.6B, and IEITYuan/Yuan-embedding-2.0-en, then ground answers with a text generation model.
  • Vision and multimodal understanding: Run image description, visual Q&A, and diagram analysis with models such as Qwen/Qwen2.5-VL-72B-Instruct and kimi-k2.6.
  • Speech applications: Build real-time transcription with bosonai/asr and conversational voice output with bosonai/tts for voice assistants and accessibility tools.
  • Long-context workflows: Process large documents or codebases in a single request with million-token models such as DeepSeek-V4-Flash.

​
Billing and Cost Optimization

  • Pay as you go: Pay only for tokens used, with stable performance from development through production.
  • Model mix: Start with lower-cost models to validate use cases, then switch to stronger models as needed.
  • Per-key limits: Set quotas and TPM/RPM limits on each API key to reduce the risk of runaway usage.

​
Services and Support

  • Enterprise support: Solution co-design, performance tuning, and production go-live assistance.
  • Migration and compatibility: OpenAI-style APIs compatible with mainstream SDKs and AI coding tools.