Update README.md (ModelCloud#46)

* Update README.md

* Update README.md

README.md

@@ -16,6 +16,32 @@
 
 *For more histories please turn to [here](docs/NEWS_OR_UPDATE.md)*
 
+
+## How is AutoGPTQ-NEXT different from AutoGPTQ?
+
+AutoGPTQ-NEXT is an updated version of AugtoGPTQ with latest bug fixes applied, new features, better/latest model support, and an guranteed from the ModelCloud.ai team and that we, along with the open-source ML community, will take every effort to bring the library up-to-date with latest advancements, model support, and bug fixes.
+
+## Mission Statement
+
+We want AutoGPTQ-NEXT to be highy focused on GPTQ based quantization and target inference compatibility with Transformers, vLLM, and SGLang. 
+
+## Major Changes vs AutoGPTQ
+
+* `Sym=False` Support. AutoGPTQ main has broken `sym=false`.
+* `lm_head` module quantized inference support for further vram reduction.
+* PENDING `lm_head` quantization will be added soon with support from Intel/Autoround.
+* ChatGLM Model Support.
+* Better defaults resulting in faster inference.
+* Better default PPL without with tweaked internal code (Result may vary depending on calibration set and gpu usage).
+* PENDING: DBRX Model Support.
+* Removed non-working, partially working, or fully deprecated features: Peft, ROCM, AWQ Gemm execution via GPTQ kernels, Triton v1 (replaced by v2).
+* Fixed Packing Performance regression on high core-count systems.
+* Thousands of lines of refractor/cleanup. 
+* Complete tests with every feature and model tested. Everything that does not pass tests will be removed from repo. We want quality over quantity.
+
+## Platform Support
+AutoGPTQ-NEXT is currently Linux only and requires Torch/Cuda capable GPU from NVIDIA. WSL on Windows should work as well. ROCM/AMD support will be re-added in a furture version after everything on ROCM has been validated. Only fully validated features will be re-added from the original AutoGPTQ repo. 
+
 ## Installation
 
 AutoGPTQ-NEXT is available for Linux only. You can install the latest stable release of AutoGPTQ from pip with pre-built wheels:
@@ -29,14 +55,12 @@ On NVIDIA systems, AutoGPTQ-NEXT does not support [Maxwell or lower](https://qii
 
 ### Install from source
 
-Clone the source code:
+Clone repo:
 ```bash
 git clone https://github.com/Qubitium/AutoGPTQ-NEXT.git && cd AutoGPTQ
 ```
 
-A few packages are required in order to build from source: `pip install numpy gekko pandas`.
-
-Then, install locally from source:
+Compile:
 ```bash
 pip install -vvv --no-build-isolation -e .
 ```