Nexa AIStreaming conversations
Ember begins responding as tokens are generated, making Nexa feel immediate and interactive instead of waiting for a full response to complete.
First Nexa model
Ember 0.5
Nexa's first-generation text model for fast assistance, early reasoning, streaming chat, and agentic workflows.
Built for Nexa's private assistant stack
Ember 0.5 is the first official model release in the Nexa model family. It powers the core Nexa chat experience and lays the foundation for browser assistance, memory-aware workflows, and permission-safe agent actions.
Release
Ember 0.5
Model type
Text / chat
Runtime
Nexa Private Runtime
Context window
Up to 262K tokens
Max output
8,192 tokens
Serving modes
Fast, Thinker, Deep Thinker
What Ember 0.5 does
Ember 0.5 focuses on making Nexa useful from day one: clear responses, streaming output, practical reasoning, and a stable foundation for AI-native browser workflows.
Everyday assistance
Built for clear answers, writing support, explanations, planning, summarization, and practical decision support.
Early reasoning
Handles multi-step prompts, coding guidance, comparisons, structured plans, and workflow breakdowns with a balanced response style.
Browser-aware workflows
Designed to support Nexa Browser features such as page understanding, form assistance, research help, and task guidance.
Tool routing foundation
Structured to support future permission-aware tool use across chat, memory, browser actions, workspace automation, and creative workflows.
Private inference first
Served through Nexa's private model runtime with CUDA-aware local inference, keeping the core assistant stack under Nexa control.
Access and usage
Ember 0.5 is available across Nexa plans with different usage limits, modes, and product capabilities.
Free
Core Ember chat access with starter usage limits.
Plus
Higher message limits, memory features, and richer assistant access.
Pro
Expanded usage for research, coding guidance, strategy, planning, and creative workflows.
Business
Team-ready access with admin controls, analytics, shared workspace direction, and private assistant features.
Runtime design
Ember 0.5 is served through the Nexa Private Runtime, not a public placeholder model page.
Mode-based generation
Nexa routes chat through Ember 0.5 for streaming responses, token tracking, and mode-based generation settings. Fast keeps answers lightweight, Thinker gives more room for planning, and Deep Thinker is reserved for heavier reasoning tasks.
Fast
Short, responsive answers for everyday tasks.
Thinker
Balanced reasoning, structure, and planning.
Deep Thinker
Longer context and deeper work.
Next improvements
- Improve first-token latency and warm runtime behavior.
- Strengthen code, research, and browser-task reliability.
- Expand tool routing with permission-aware browser actions.
- Tune response style so Nexa stays direct, professional, and consistent.
- Prepare Ember 1.0 as the next larger production model release.
- Evolve Prism with stronger prompt following, sharper detail, and richer creative controls.
Image model
Prism 0.5
Prism 0.5 is Nexa's first-generation image model, built for visual creation, concept art, image ideation, and creative workflows.
What Prism 0.5 does
Prism gives Nexa a dedicated visual model identity for generated images, creative direction, and image-library workflows.
Release
Prism 0.5
Model type
Image generation
Runtime
Nexa Image Runtime
Input
Natural-language prompts
Output
Generated image assets
Focus
Visual creation and ideation
Text-to-image creation
Turns natural-language prompts into generated visuals for concepts, mockups, creative drafts, and design exploration.
Concept direction
Built for product concepts, brand visuals, app hero images, environment ideas, and early creative exploration.
Prompt refinement
Works with Nexa chat so users can describe, revise, and continue visual ideas without leaving the workspace.
Library integration
Generated images appear in the user's Nexa image library so they can review, reopen, and manage previous outputs.