Nano
350MA 350M dense tier for fast iteration and end-to-end validation. It uses 18 Mamba-3 blocks plus 6 MLA blocks, all parameters active, and no MoE routing.
Total Parameters350M
Active Parameters350M
Base Context4k
Long Context8k
Ivonar: Native 1-Bit LLM Family.
Ivonar is a native 1-Bit LLM family training real W1.58A8 BitLinear models in PyTorch with ternary-weight STE, INT8-activation STE, Mamba-3 mixers, MLA attention, and dense-to-sparse scaling across four tiers.
About Ivonar
Ivonar is a native 1-Bit LLM family developed independently in Germany. It spans four training scales: Nano, Mini, Medium, and High. Each tier shares the same technical direction while scaling model capacity, active parameters, context length, and dense-to-sparse architecture choices.
Nano / Mini / Medium / High
Mini
1.5BA compact 1.5B dense model for the first full production scale. It keeps all parameters active, combines Mamba-3 with MLA, and avoids expert routing.
Total Parameters1.5B
Active Parameters1.5B
Base Context8k
Long Context64k
Medium
8.0B / 3.0B activeAn 8.0B total parameter tier with 3.0B active parameters. Sparse capacity is limited to feedforward blocks with Top-2 routing; attention and sequence mixers stay shared.
Total Parameters8.0B
Active Parameters3.0B active
Base Context16k
Long Context128k
High
32.1B / 6.3B activeA 32.1B total parameter flagship tier with 6.3B active parameters. It extends the FFN-only sparse MoE pattern with Top-2 routing and the longest context target.
Total Parameters32.1B
Active Parameters6.3B active
Base Context32k
Long Context256k
Team / About Us
Focused independent AI engineering from Germany.
Ivonar was founded and is led by Luis Oezdem. We operate as an independent research and engineering initiative focused on efficient 1-Bit LLM development. Architecture, training infrastructure, data pipeline design, and deployment direction are kept tightly aligned under one technical vision.
The operating model is deliberately execution-oriented: validate the system through Ivonar Nano, then scale carefully into Mini, Medium, and High with clear quality, efficiency, and deployment goals.
ResearchEngineeringDeployment
Architecture & Core Features
The 1-Bit stack is built around real W1.58A8 BitLinear training in PyTorch: ternary-weight STE, INT8-activation STE, Mamba-3 state-space mixers, MLA latent attention, and FFN-only sparse routing where the larger tiers need it. Packed or custom kernels are future inference optimizations, not the current training GEMM.
W1.58A8 BitLinear Training
Ivonar trains real BitLinear models with ternary-weight STE and INT8-activation STE in PyTorch. It is not yet a custom bit-packed CUDA training GEMM; packed/custom kernels are future inference work.
Mamba-3 and MLA Hybrid
Mamba-3 state-space mixers handle efficient long-range sequence processing. MLA latent attention provides retrieval-focused attention, with Nano explicitly using 18 Mamba-3 blocks and 6 MLA blocks.
Dense and Sparse Tiers
Nano and Mini are dense models with no MoE. Medium and High use FFN-only sparse MoE with Top-2 routing, keeping sparse capacity out of attention and sequence-mixing blocks.
Long Context with LongRoPE2
Context targets scale from 4k base and 8k long in Nano up to 32k base and 256k long in High, using separate base and long-context continuation phases.
Efficiency / Target Audience
A powerful European 1-Bit model built for efficient use.
Ivonar aims to become the leading European 1-Bit LLM: real W1.58A8 BitLinear training in PyTorch today, honest dense and sparse tiering, and practical deployment work that can later benefit from packed or custom inference kernels.
Strong results with compact models
Ternary BitNet layers achieve competitive language, code, and math capability without dense-parameter scaling. The goal is practical quality across daily work, prototypes, and research tasks.
Individuals, builders, research workflowsMinimal memory footprint
Ternary weights, Mamba-3 mixers, and dense-to-sparse scaling keep active compute controlled. Nano and Mini are dense; Medium and High use FFN-only sparse MoE with Top-2 routing.
Practical inference and controlled scalingBuilt for everyone.
Ivonar serves both B2C and B2B markets. Our monetization strategy relies on commercial API access for developers, premium subscriptions for end-users, and flexible enterprise licensing for businesses that require strict data privacy and local deployment.
Flexible access for users, builders, and teamsRoadmap
Architecture defined. Nano training is active. Mini, Medium, and High scale from there.
01
DoneArchitecture
Completed
Hybrid 1-Bit architecture with PyTorch W1.58A8 BitLinear training, Mamba-3 mixers, MLA attention, dense Nano/Mini tiers, and FFN-only sparse MoE for Medium/High is defined.
02
ActiveNano Training
In Training
Training the 350M Nano tier with 4k base and 8k long context for fast iteration, pipeline validation, and end-to-end BitLinear quality checks.
03
NextMini Training
Upcoming
Next dense tier: 1.5B total and active parameters, 8k base context, 64k long context, Mamba-3 + MLA, and no MoE routing.
04
PlannedSparse Scale-Up
Upcoming
Medium and High add FFN-only sparse MoE with Top-2 routing: 8.0B total / 3.0B active for Medium, then 32.1B total / 6.3B active for High.