Est. 2026-2028
Federated Learning
Distributed training across edge devices.
AI Market · Lineage
The LLM evolutionary tree.
A living lineage of the frontier and open-weight models the weekly record is scoring against.
Architectural branches
- Foundational
- Encoder-only
- Encoder-decoder
- Decoder-only
- Mixture-of-Experts
- Multimodal
- Reasoning
Future horizon
Directional hypotheses, not forecasts.
Est. 2026-2028
Continual Learning
Online weight updates without catastrophic forgetting.
Est. 2026-2028
Sparse Models
Native sparse computation beyond MoE routing.
Est. 2027-2029
Neuromorphic Models
Spiking-neuron architectures on neuromorphic chips.
Est. 2026-2027
Energy-Efficient Models
FP4 / 1-bit weights, in-memory compute primitives.
Explore the tree
34 models in this view
100%
Drag to pan · Scroll or pinch to zoom · Select a model to trace its lineage and inspect the record. Each column is one lab, so a family's newest model sits above its predecessor. Branch color marks architectural evolution; solid lines are direct parents and dashed lines are influences. Filled nodes are open source or open weights. Dashed outlines mark lower-confidence placement.
No model selected.
Companion systems
261 models with cited sources, infrastructure signals, and explicit confidence levels.