Model Roadmap
Planned ArtixCode models — from smartphone SLMs to frontier research
Three model families with multiple planned variants for coding, debugging, analysis, vulnerability assistance, codebase audit, and Bangla-English developer workflows.
Roadmap, not hype
ArtixCode's model family is planned in stages. Smaller SLM and local models are the practical first step. Larger parameter targets represent long-term research direction and may change based on infrastructure, data quality, safety evaluation, and product needs.
These models are part of the ArtixCode roadmap. Parameter targets, capabilities, and release timelines may change as research, infrastructure, and safety evaluation progress.
Model families & variants
Each family includes planned variants from lightweight SLMs to long-term frontier research targets. None are production-ready unless explicitly marked available.
Mohuya family
Planned coding assistance models for code generation, explanation, refactoring, test generation, and Bangla-English developer education.
Primary focus: Coding assistance, code generation, code explanation, refactoring, test generation, Bangla-English developer education.
Mohuya Nano
Target: 0.5B-1B parameters
A planned small language model for lightweight coding help, short explanations, and offline developer education.
Target devices
Focus areas
- Basic coding help
- Short code explanation
- Bangla-English learning support
- Simple refactoring
- +1 more planned focus areas
Availability
Planned roadmap — not currently available
Mohuya Lite
Target: 3B parameters
A planned mobile-friendly coding model for everyday programming help and lightweight code generation.
Target devices
Focus areas
- Code completion
- Simple debugging
- Code explanation
- Bangla-English developer support
Availability
Planned roadmap — not currently available
Mohuya Coder
Target: 7B parameters
A planned practical coding model for local developer workflows, code generation, debugging, explanation, and refactoring.
Target devices
Focus areas
- Code generation
- Refactoring
- Test generation
- Bangla-English explanations
- +1 more planned focus areas
Availability
Upcoming — planned for local developer workflows
Mohuya Pro
Target: 15B parameters
A planned stronger coding model for deeper software engineering assistance, better reasoning, and more reliable multi-file understanding.
Target devices
Focus areas
- Multi-step code generation
- Debugging
- Framework-specific reasoning
- Test planning
- +1 more planned focus areas
Availability
Research planned — cloud inference target
Mohuya Max
Target: 30B parameters
A planned advanced coding model for larger context, deeper reasoning, and professional software workflows.
Target devices
Focus areas
- Large-context coding help
- Advanced refactoring
- Code architecture support
- Framework migration planning
Availability
Research planned — professional cloud target
Mohuya Frontier
Long-term target: 500B parameters
A long-term research target for a frontier-scale ArtixCode coding model focused on multilingual software engineering.
Target devices
Focus areas
- Advanced coding intelligence
- Multilingual developer support
- Large-scale software reasoning
- Autonomous coding workflows
Availability
Private research — long-term frontier target
From smartphone SLMs to frontier research
Planned scale progression across all ArtixCode model families. Smaller variants are practical first steps; frontier targets represent long-term research direction.
Nano
0.5B–1B
Smartphone-runnable SLM targets for lightweight help
Lite
3B
Mobile/on-device targets for everyday developer workflows
Core
7B
Local developer models — Coder, Debug, Review
Pro
15B
Cloud professional models — Pro, Analyst, Audit
Max
30B
Enterprise-scale models — Max, Secure, Enterprise
Frontier
500B–700B
Long-term research targets — not currently available
Capability matrix
Planned capability focus across model families. Values reflect roadmap intent—not current availability or benchmark results.
| Capability | Mohuya | Aziha | Aronna |
|---|---|---|---|
| Code generation | Planned | Not primary focus | Not primary focus |
| Debugging | Planned | Advanced planned | Research planned |
| Bangla-English explanation | Planned | Planned | Not primary focus |
| Architecture analysis | Planned | Advanced planned | Research planned |
| Vulnerability assistance | Not primary focus | Advanced planned | Research planned |
| Codebase audit | Not primary focus | Planned | Advanced planned |
| Repository scanning | Not primary focus | Planned | Advanced planned |
| Enterprise security review | Not primary focus | Research planned | Advanced planned |
| Autonomous developer workflows | Research planned | Research planned | Advanced planned |
Built for defensive, authorized engineering
- ArtixCode models are intended to help developers improve code quality, reliability, and security.
- Vulnerability assistance must be used only on owned or authorized systems.
- Outputs must be reviewed before production use.
- Private user code will not be used for model training without explicit consent.
- Official benchmark results will be published only after evaluation.
Roadmap timeline
ArtixCode's phased approach from web assistant to planned frontier research models.
Phase 1
ArtixCode web assistant
Live assistant workflows for generation, debugging, and explanation on the ArtixCode platform.
Phase 2
Bangla-English coding dataset
Curated Bangla-English coding dataset with consent-first collection for multilingual developer support.
Phase 3
SLM and mobile targets
Research planning for Nano and Lite variants — planned smartphone-runnable SLMs and mobile/on-device coding models.
Phase 4
Local 7B developer models
Upcoming local models — Mohuya Coder, Aziha Debug, and Aronna Review for practical developer workstations.
Phase 5
Cloud 15B/30B professional models
Research planning for Pro, Analyst, Audit, Max, Secure, and Enterprise variants for team and cloud inference.
Phase 6
Frontier research targets
Long-term private research for 500B–700B frontier models across Mohuya, Aziha, and Aronna families.
Ready to shape the future of coding?
Join the ArtixCode private beta and help build Bangladesh-born AI coding intelligence for developers worldwide.