Update README.md

README.md

@@ -1,53 +1,117 @@
 ---
+language:
+- en
 license: apache-2.0
 tags:
-  - chimera
-  - moe
-  - mixture-of-experts
-  - gguf
-  - klyrone
-language:
-  - en
+- mixtral
+- moe
+- mixture-of-experts
+- merged
+- chimera
+- klyrone
+- instruct
+- text-generation
+base_model:
+- mistralai/Mixtral-8x7B-v0.1
+- mistralai/Mixtral-8x7B-Instruct-v0.1
+model_type: mixtral
 pipeline_tag: text-generation
+library_name: transformers
+inference: true
 ---
 
 # Chimera 47B
 
-**Chimera** is a Mixture-of-Experts language model developed by [Klyrone F.Z.E](https://klyrone.com), built using our proprietary **Amalgamation of Experts (AoE)** technique. Chimera features 8 specialized expert networks with top-2 routing, delivering strong instruction-following and reasoning capabilities with an efficient 12.9B active parameter footprint.
+**Klyrone F.Z.E.** · March 2026 · Apache 2.0
 
-## Model Details
+Chimera 47B is a 46.7B parameter Mixture-of-Experts language model built using Klyrone's MoE assembly framework. It delivers instruction-following, code generation, and reasoning at 154 tokens/second on H200 hardware, with only 12.9B parameters active per token.
+
+A technical paper detailing the methodology is forthcoming.
+
+---
+
+## Key Numbers
 
 | | |
 |---|---|
-| **Architecture** | Mixture of Experts (MoE) - 8 experts, top-2 routing |
-| **Total Parameters** | 46.7B |
-| **Active Parameters** | 12.9B per token |
-| **Context Length** | 32,768 tokens |
-| **Quantization** | Q5_K_M (GGUF) |
-| **Developed by** | Klyrone F.Z.E. |
+| Total Parameters | 46.7 B |
+| Active / Token | 12.9 B |
+| Architecture | MoE · 8 experts · top-2 routing |
+| Context Length | 32,768 tokens |
+| Generation Speed | 154 t/s · H200 |
+| Prompt Processing | 878 t/s · H200 |
+| Quantization | Q5_K_M · 5.69 BPW |
+| File Size | 33.2 GB GGUF |
+| License | Apache 2.0 |
+
+---
+
+## Capabilities
 
-## Key Features
+- ✅ Instruction following — multi-turn conversational coherence
+- ✅ Code generation — correct, edge-case-aware output
+- ✅ Creative writing — long-form prose and poetry
+- ✅ Factual reasoning — physics, mathematics, general knowledge
+- ✅ Consumer-grade deployment — fits accessible GPU budgets at Q5_K_M
 
-- **Efficient MoE Architecture** - Only 12.9B parameters active per forward pass despite 46.7B total, enabling fast inference
-- **Specialized Expert Networks** - 8 expert FFN modules with learned routing for task-adaptive computation
-- **Instruction-Tuned Experts** - Expert networks optimized for instruction following, code generation, and reasoning
-- **Long Context** - Supports up to 32K token context windows with RoPE positional encoding
+> Formal benchmark results (MMLU, HellaSwag, ARC-Challenge, GSM8K) in progress.
+
+---
 
-## Amalgamation of Experts (AoE)
+## About the Approach
 
-Chimera is built using our **AoE** technique, a novel approach to constructing high-quality MoE models by strategically assembling expert networks. AoE enables the creation of models that combine specialized capabilities from multiple training paradigms into a unified, coherent architecture.
+Klyrone's MoE assembly framework constructs high-performance models by composing expert sub-networks from compatible source models — without full retraining. The approach preserves routing coherence while inheriting specialized capabilities from donor models.
+
+Full methodology will be published via arXiv. For enterprise licensing or research collaboration, contact **research@klyrone.com**
+
+---
 
 ## Usage
 
-### With llama.cpp
+### llama.cpp
+
 ```bash
-./llama-cli -m Chimera-8x7B-Q5_K_M.gguf -p "Your prompt here" -n 500 -ngl 99
+./llama-cli \
+  -m Chimera-47B-Q5_K_M.gguf \
+  -p "You are a helpful assistant." \
+  --ctx-size 32768 \
+  -n 512
 ```
 
-## Intended Use
+### Transformers
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained(
+    "klyrone/Chimera",
+    device_map="auto",
+    torch_dtype="auto"
+)
+tokenizer = AutoTokenizer.from_pretrained("klyrone/Chimera")
+```
+
+---
+
+## Limitations
 
-Chimera is designed for general-purpose text generation including conversational AI, code generation, reasoning, and instruction following.
+- Router fine-tuning not yet applied — a short gate re-alignment is expected to yield marginal quality gains
+- No independent safety evaluation conducted — not recommended for unsupervised public-facing deployment
+- Benchmark results pending publication
 
-## License
+---
+
+## Citation
+
+```bibtex
+@misc{chimera47b2026,
+  title        = {Chimera 47B},
+  author       = {{Klyrone F.Z.E.}},
+  year         = {2026},
+  howpublished = {\url{https://huggingface.co/klyrone/Chimera}}
+}
+```
+
+---
 
-Apache 2.0
+*Chimera 47B · Klyrone F.Z.E. · Apache 2.0 · A technical paper on the AoE technique is forthcoming.*