temaq-org/Tema_Q-X4-8B-Thinking

Partial fine-tune vs google/gemma-4-E4B-it. Partial change: 10% of the weights changed (attention, mlp, other), in 27 of 42 layers.

Tensors changed: MLP 23/606, Norms 0/502, other 24/261, Attention 27/755, Embeddings 0/6

3 derivatives checked

• mradermacher/Tema_Q-X4-8B-Thinking-i1-GGUF: Quantization (fewer layers)
• mradermacher/Tema_Q-X4-8B-Thinking-GGUF: Same weights, converted (fewer layers)
• TaQuants/Tema_Q-X4-8B-Thinking-TaQuants-GGUF: Quantization (fewer layers)

On the Hub · Downloads on Model Pulse