diff --git a/gallery/index.yaml b/gallery/index.yaml
index 059f3503871d..4e994e940907 100644
--- a/gallery/index.yaml
+++ b/gallery/index.yaml
@@ -22599,3 +22599,51 @@
     - filename: Almost-Human-X3-32bit-1839-6B.i1-Q4_K_M.gguf
       sha256: 5dc9766b505d98d7a5ad960b321c1fafe508734ca12ff4b7c480f8afbbc1e03b
       uri: huggingface://mradermacher/Almost-Human-X3-32bit-1839-6B-i1-GGUF/Almost-Human-X3-32bit-1839-6B.i1-Q4_K_M.gguf
+- !!merge <<: *jamba
+  name: "ai21-jamba-reasoning-3b"
+  urls:
+    - https://huggingface.co/Mungert/AI21-Jamba-Reasoning-3B-GGUF
+  description: |
+    **Model Name:** AI21 Jamba Reasoning 3B
+    **Repository:** [ai21labs/AI21-Jamba-Reasoning-3B-GGUF](https://huggingface.co/ai21labs/AI21-Jamba-Reasoning-3B-GGUF)
+    **License:** Apache 2.0
+    **Pipeline:** Text Generation
+    **Architecture:** Hybrid Transformer–Mamba (28 layers: 26 Mamba, 2 Attention)
+    **Parameters:** 3B
+    **Context Length:** Up to 256,000 tokens
+    **Quantizations Available:** FP16 (6.4 GB), Q4_K_M (1.93 GB)
+    **Developed by:** AI21
+
+    ---
+
+    ### 🔍 **Overview**
+    AI21’s Jamba Reasoning 3B is a compact, high-performance reasoning model that combines the strengths of Transformers and Mamba architecture for efficient long-context processing. Despite its small size, it delivers top-tier intelligence scores—surpassing models like Llama 3.2 3B and Gemma 3 4B—on benchmarks such as MMLU-Pro and IFBench.
+
+    ### ⚡ **Key Features**
+    - **Ultra-long context (256K tokens):** Mamba layers enable scalable, memory-efficient handling of extremely long inputs.
+    - **High efficiency:** Optimized for edge devices, laptops, and low-resource environments without sacrificing performance.
+    - **Strong reasoning capabilities:** Fine-tuned for structured reasoning, tool use, code generation, and complex problem solving.
+    - **Multiple runtime support:** Works with `llama.cpp`, `vLLM`, `Ollama`, and `LM Studio`.
+
+    ### 📊 **Performance Highlights**
+    - **MMLU-Pro:** 61.0%
+    - **Humanity’s Last Exam:** 6.0%
+    - **IFBench:** 52.0%
+
+    > ✅ Ideal for use cases requiring fast, accurate reasoning over long documents—e.g., document analysis, code generation, customer support ticket triage, and research summarization.
+
+    ### 📥 **Quick Start**
+    Use with `llama.cpp` or `vLLM` for local inference. Pre-quantized GGUF files are available for lightweight deployment.
+
+    > 📌 **Note:** This is the *original* model from AI21. The repository `Mungert/AI21-Jamba-Reasoning-3B-GGUF` is a user-quantized version; always refer to `ai21labs/AI21-Jamba-Reasoning-3B-GGUF` for the official, unquantized model description.
+
+    ---
+
+    🔗 **Learn More:** [AI21 Blog – Introducing Jamba Reasoning 3B](https://www.ai21.com/blog/introducing-jamba-reasoning-3B)
+  overrides:
+    parameters:
+      model: AI21-Jamba-Reasoning-3B-q4_k_m.gguf
+  files:
+    - filename: AI21-Jamba-Reasoning-3B-q4_k_m.gguf
+      sha256: a5ad6704bb679010ffeea5a2a8c06e652ce7dec0c955f8945586ba2f5a643cc7
+      uri: huggingface://Mungert/AI21-Jamba-Reasoning-3B-GGUF/AI21-Jamba-Reasoning-3B-q4_k_m.gguf