CosmicFish Models

Complete lineup of AI models from compact to advanced.

Language Models

CosmicFish Pico

Research Preview

Compact 19M parameter model powered by our custom-built Cosmicizer tokenizer for efficient performance.

Parameters: 19 Million
Use Case: Ultra-efficient tasks

CF30M

Research Preview

Lightweight model perfect for resource-constrained environments and testing.

Parameters: 30 Million
Use Case: IoT, edge computing

CF90M

Available Now

Balanced model offering decent performance for everyday AI tasks and applications.

Parameters: 90 Million
Use Case: General conversation

CF120M

Available

Bigger model delivering excellent performance across all AI tasks with optimal efficiency.

Parameters: 120 Million
Use Case: Content generation, Q&A

CF300M

Available

Our most advanced language model with superior capabilities in coding, mathematics, and reasoning.

Parameters: 300 Million
Use Case: Code, math, reasoning

Specialized Models

Purpose-built models for specific applications

CosmicGen

In Development

Advanced image generation model for high-quality visual content creation.

Type: Image Generation
Output: 256x256 images
Use Case: Art creation, design

CosmicFish MOE

In Development

Mixture of Experts architecture for enhanced performance and efficiency.

Architecture: Mixture of Experts
Experts: 2x70M parameters
Use Case: Specialized tasks

Model Comparison

Compare specifications across our model lineup

Model Parameters Model Size (Un-quantized) Memory Status
CosmicFish Pico 19M ~220 MB ~512 MB Research Preview
CF30M 30M ~370 MB ~1 GB Research Preview
CF90M 90M ~1 GB ~2 GB Available
CF120M 120M ~1.4 GB ~2 GB Available
CF300M 369M ~4 GB ~6 GB Available