Models

Eleven frontier models, one API

From flagship reasoning to real-time voice — explore every model in the Somix family. All accessible through a single OpenAI-compatible API.

Text Models

Flagship

Somix Prime 5.2

somix-prime-5.2

Our most capable reasoning model.

Trillion-parameter MoE architecture with 42B active parameters and 1M-token context window. State-of-the-art reasoning, coding, and agentic performance matching the best frontier models.

Capabilities

  • 1M-token ultra-long context
  • Advanced reasoning & math
  • Agentic tool-use
  • Multilingual understanding
  • Complex code generation
Try Somix Prime

Specifications

Total parameters1T (MoE)
Active parameters42B
Context window1,000,000 tokens
Input pricing$3.00 / MTok
Output pricing$6.00 / MTok
View API docs
Fast

Somix Swift 5.1

somix-swift-5.1

Blazing fast, production-ready.

FP4 hybrid quantization combined with DFlash speculative decoding and TileRT kernel optimization delivers groundbreaking 1000 tokens/s throughput for real-time applications.

Capabilities

  • 1000 tokens/s peak throughput
  • Real-time chat latency
  • High concurrency support
  • Standard GPU deployment
  • Streaming responses
Try Somix Swift

Specifications

Peak speed1000 tokens/s
QuantizationFP4 hybrid
Context window128K tokens
Input pricing$1.00 / MTok
Output pricing$2.00 / MTok
View API docs

Specialized Models

Developer

Somix Code 5.2

somix-code-5.2

Your AI pair programmer.

Next-generation AI coding assistant with unlimited context. Built-in persistent memory that understands your entire codebase, your style, and your conventions.

Capabilities

  • Unlimited context window
  • Persistent project memory
  • VS Code & JetBrains
  • SWE-Bench Pro top-tier
  • Code review & refactoring
Try Somix Code

Specifications

Context windowUnlimited
IDE supportVS Code, JetBrains, Neovim
BenchmarksSWE-Bench Pro top tier
Included inPro plan ($20/mo)
Languages80+ programming languages
View API docs
Emotional AI

Somix Persona 5.0

somix-persona-5.0

Emotional AI companions.

Dedicated model for virtual companions, IP avatars, and celebrity personas. Deep storyline progression with persona consistency and safety guardrails.

Capabilities

  • 97% persona fidelity
  • Group chat mode
  • Function calling
  • Safety & compliance
  • IP avatar customization
Try Somix Persona

Specifications

Persona fidelity97% (benchmark)
Group chatSupported
Dialogue depthLong-term memory
Free tierAvailable
IP licensingContact us
View API docs

Vision & Media

Vision

Somix Vision 5.1

somix-vision-5.1

See and understand.

Advanced visual understanding with OCR, chart interpretation, document analysis, and multi-image reasoning. Process images, screenshots, and diagrams with high accuracy.

Capabilities

  • OCR & chart understanding
  • Document analysis
  • Multi-image reasoning
  • Screenshot interpretation
  • Diagram parsing
Try Somix Vision

Specifications

Max image resolution8K
Multi-imageUp to 20 images
OCR accuracy99%+
Input pricing$3.00 / MTok
Image input$0.005 / image
View API docs
Video

Somix Video 5.0

somix-video-5.0

Generate video from text.

Video generation 2.0 with sub-200ms audio-synced generation. Ideal for live streaming, real-time ads, and dynamic content creation at scale.

Capabilities

  • <200ms generation latency
  • Audio-synced output
  • Live streaming support
  • Real-time ad generation
  • Dynamic content creation
Try Somix Video

Specifications

Generation latency< 200ms (fast)
Max duration60 seconds
ResolutionUp to 1080p
Audio syncBuilt-in
PricingPer-second
View API docs
Image

Somix Image 5.1

somix-image-5.1

Create stunning visuals.

High-quality image generation with precise prompt following. Supports text-to-image, image-to-image, inpainting, outpainting, and style transfer.

Capabilities

  • Text-to-image generation
  • Image-to-image editing
  • Inpainting & outpainting
  • Style transfer
  • Precise prompt following
Try Somix Image

Specifications

Max resolution4K
Generation speed< 3 seconds
Aspect ratiosAny
Styles100+ presets
PricingPer-image
View API docs

Audio & Voice

Real-time Voice

Somix Live 5.2

somix-live-5.2

Real-time voice conversations.

Real-time voice dialogue with sub-300ms response latency. Natural conversation flow with interruption handling, emotion detection, and multi-turn context.

Capabilities

  • <300ms response latency
  • Interruption handling
  • Emotion detection
  • Multi-turn context
  • WebSocket streaming
Try Somix Live

Specifications

Response latency< 300ms
ProtocolWebSocket
Emotion detection8 emotions
Languages12 supported
PricingPer-minute
View API docs
Translation

Somix Translate 5.1

somix-translate-5.1

Real-time simultaneous interpretation.

Low-latency streaming translation across 12 languages. Perfect for multilingual meetings, live content localization, and cross-border communication.

Capabilities

  • 12 language pairs
  • Real-time streaming
  • Simultaneous interpretation
  • Meeting integration
  • Live captioning
Try Somix Translate

Specifications

Languages12
Translation latency< 500ms
ModeStreaming
Accuracy (BLEU)42+
PricingPer-minute
View API docs
TTS

Somix Voice 5.0

somix-voice-5.0

Natural text-to-speech.

High-fidelity text-to-speech with 4.7 MOS naturalness. Voice cloning from just a 3-second sample. Expressive, emotional speech synthesis with 50+ voices.

Capabilities

  • 4.7 MOS naturalness
  • 3-second voice cloning
  • 50+ preset voices
  • Emotional expression
  • SSML support
Try Somix Voice

Specifications

MOS score4.7
Voice cloning3-second sample
Preset voices50+
Output formatsMP3, WAV, Opus
PricingPer-character
View API docs
Music

Somix Music 5.0

somix-music-5.0

Compose music with AI.

AI music generation from text prompts. Compose original tracks, background scores, and soundscapes in diverse genres — from classical to electronic.

Capabilities

  • Text-to-music generation
  • Multi-genre support
  • Full-length tracks
  • Background scoring
  • Style mixing
Try Somix Music

Specifications

Max duration5 minutes
Genres50+
Output quality44.1 kHz stereo
Stems exportComing soon
PricingPer-track
View API docs