Mercury Edit 2
AI model · Code model · Text → Code by Inception Labs
Ultra-fast diffusion LLM for code autocomplete and next-edit suggestions.
Mercury Edit 2 is a diffusion-based large language model specifically engineered for high-speed code generation and editing. It supports two primary developer workflows: Fill-In-the-Middle (FIM) for intelligent autocomplete and Next-Edit for suggesting incremental code improvements. By leveraging Inception's diffusion architecture, the model provides ultra-fast responses that allow developers to stay "in-the-flow" without waiting for sequential token generation. It is designed to be integrated into IDEs and code editors to power responsive autocomplete and intelligent tab suggestions. The mode
- Languages
- English
- Context Window (FIM)
- 32,000 tokens
- Context Window (NextEdit)
- 32,000 tokens
- Max Output Tokens
- 8,192 tokens
- Input Modality
- Text
- Output Modality
- Text
Overview
Mercury Edit 2 is a diffusion-based large language model specifically engineered for high-speed code generation and editing. It supports two primary developer workflows: Fill-In-the-Middle (FIM) for intelligent autocomplete and Next-Edit for suggesting incremental code improvements. By leveraging Inception's diffusion architecture, the model provides ultra-fast responses that allow developers to stay "in-the-flow" without waiting for sequential token generation. It is designed to be integrated into IDEs and code editors to power responsive autocomplete and intelligent tab suggestions. The model handles a 32,000 token context window to maintain awareness of relevant code snippets and file history.
Coding Workflows
- Autocomplete (FIM): Provides ultra-fast code generation using the Fill-In-the-Middle (FIM) paradigm.
- Next-Edit Suggestions: Suggests the next logical edit in a codebase based on current file content and cursor position.
- 32K Context Window: Supports a 32,000 token context window for both FIM and NextEdit tasks.
Integration
- Low-Latency Editing: Designed to deliver code suggestions instantly to keep developers in the flow.
- Developer Tool Integration: Accessible via dedicated completions endpoints for seamless IDE integration.
Pricing
- Input
- $0.25 per 1M tokens
- Output
- $0.75 per 1M tokens
- Cached Input
- $0.025 per 1M tokens
Specifications
- Context Window (FIM)
- 32,000 tokens
- Context Window (NextEdit)
- 32,000 tokens
- Max Output Tokens
- 8,192 tokens
- Input Modality
- Text
- Output Modality
- Text
Added by AntSpace from public sources (docs.inceptionlabs.ai).