Eunoia-4B-Mini
An open-source foundation model for long-horizon reasoning.
Eunoia-4B-Mini is the first open-source model released under the Eunoia research program by SHV Groups. It is designed to explore how reasoning unfolds when intelligence is constrained by efficiency rather than scale.
Instead of pursuing parameter count as a proxy for intelligence, Eunoia-4B-Mini focuses on reasoning continuity — the ability to maintain coherent, multi-step thought across extended contexts.
Why Eunoia-4B-Mini Exists
The modern AI ecosystem is dominated by large models that perform exceptionally well on short-horizon tasks, yet often degrade when required to plan, reason, or adapt across longer sequences of thought.
Eunoia-4B-Mini challenges this trajectory. It is intentionally compact, intentionally constrained, and intentionally open — allowing researchers to study where reasoning breaks, where it persists, and how it can be improved without brute force scaling.
Design Intent & Capabilities
At approximately four billion parameters, Eunoia-4B-Mini occupies a deliberate middle ground. It is large enough to express structured reasoning, yet small enough to remain reproducible, inspectable, and adaptable by the broader research community.
The model is well-suited for experimentation in long-context reasoning, sequential decision-making, analytical tasks, and agent-based systems where persistence of thought matters more than surface fluency.
Open-Source & Responsible by Design
Eunoia-4B-Mini is released under an Apache-2.0 license as part of SHV Groups’ commitment to open research and responsible AI development.
This model is intended for research, experimentation, and further fine-tuning. Outputs should be evaluated critically, and the model should not be treated as an authoritative or autonomous decision-maker.
Eunoia-4B-Mini represents the first step in a broader research journey. It is not positioned as a final system, but as a foundation for learning how intelligence reasons, persists, and evolves under constraint.
This release marks SHV Groups’ commitment to democratizing AI research — not by hiding intelligence behind closed systems, but by building openly, carefully, and responsibly.