Update README via Huggy

README.md

@@ -67,17 +67,16 @@ Command A+ is an open source model with 25 billion active parameters and 218B to
 
 Developed by: [Cohere](https://cohere.com/) and [Cohere Labs](https://cohere.com/research)
 
-* Point of Contact: [**Cohere Labs**](https://cohere.com/research)  
-* License: [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)  
-* Model: command-a-plus-05-2026  
-* Model Size: 25B active parameters, 218B total parameters  
+* Point of Contact: [**Cohere Labs**](https://cohere.com/research)
+* License: [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
+* Model: command-a-plus-05-2026
+* Model Size: 25B active parameters, 218B total parameters
 * Context length: 128K input
 
-For more details about this model, please check out our [blog post](http://cohere.com/blog/command-a-plus). 
+For more details about this model, please check out our [blog post](http://cohere.com/blog/command-a-plus).
 
 You can try out Command A+ before downloading the weights in our hosted [Hugging Face Space](https://huggingface.co/spaces/CohereLabs/command-a-plus-05-2026).
 
-
 **Available quantizations**
 
 The following quantizations are available with example minimum GPU requirements
@@ -90,7 +89,7 @@ The following quantizations are available with example minimum GPU requirements
 
 All three quantizations show negligible differences in benchmark quality and performance. **Our recommended quantization for most uses is [W4A4](https://huggingface.co/CohereLabs/command-a-plus-05-2026-w4a4) which boasts superior speed and latency characteristics alongside a smaller hardware footprint.**
 
-For more details, please check out our [blog post](http://cohere.com/blog/command-a-plus). 
+For more details, please check out our [blog post](http://cohere.com/blog/command-a-plus).
 
 
 **Usage**
@@ -117,9 +116,9 @@ input_ids = tokenizer.apply_chat_template(
 )
 
 gen_tokens = model.generate(
-    input_ids, 
-    max_new_tokens=4096, 
-    do_sample=True, 
+    input_ids,
+    max_new_tokens=4096,
+    do_sample=True,
     temperature=0.6,
     top_p=0.95
 )
@@ -174,7 +173,7 @@ print(outputs[0]["generated_text"][-1])
 Command A+ `w4a4` can only run on `vLLM >=0.21.0`. W4A4 and accurate response parsing also requires installing Cohere’s melody library.
 
 ```sh
-uv pip install vllm>=0.21.0 
+uv pip install vllm>=0.21.0
 uv pip install transformers
 uv pip install cohere_melody>=0.9.0
 ```
@@ -188,15 +187,15 @@ vllm serve CohereLabs/command-a-plus-05-2026-w4a4 -tp 1 --tool-call-parser coher
 
 We recommend using the following set of sampling parameters for generation: `temperature=0.9`, `top_p=0.95`, `repetition_penalty=1.04`.
 
-**Quantization Methodology:** Reasoning models pay an outsized quantization tax: long decoding traces compound per-token errors, so naive low-bit conversion typically shows up as visible regressions on hard benchmarks. To mitigate this, we quantize selectively and use distillation to close the residual quality gap. We apply NVFP4 W4A4 quantization (4-bit weights and activations, with two-level scaling) to the MoE experts only. The attention path, i.e., Q/K/V/O projections, the KV cache, and attention compute, is kept at full precision. MoE experts dominate total parameter count, so quantizing them to 4 bits brings the model within the memory budget of a single B200 and accelerates the expert GEMMs that bottleneck short-to-medium-context decode. Furthermore, we use Quantization-Aware Distillation (QAD) in the post-training phase: the quantized student is trained to match the full-precision teacher's output distribution, with fake quantization operators in the forward pass and straight-through estimators on the backward. 
+**Quantization Methodology:** Reasoning models pay an outsized quantization tax: long decoding traces compound per-token errors, so naive low-bit conversion typically shows up as visible regressions on hard benchmarks. To mitigate this, we quantize selectively and use distillation to close the residual quality gap. We apply NVFP4 W4A4 quantization (4-bit weights and activations, with two-level scaling) to the MoE experts only. The attention path, i.e., Q/K/V/O projections, the KV cache, and attention compute, is kept at full precision. MoE experts dominate total parameter count, so quantizing them to 4 bits brings the model within the memory budget of a single B200 and accelerates the expert GEMMs that bottleneck short-to-medium-context decode. Furthermore, we use Quantization-Aware Distillation (QAD) in the post-training phase: the quantized student is trained to match the full-precision teacher's output distribution, with fake quantization operators in the forward pass and straight-through estimators on the backward.
 
 ## **Model Details**
 
 **Input**: Text and images.
 
-**Output**: Model generates text. 
+**Output**: Model generates text.
 
-**Model Architecture**: Command A+ is a decoder-only Sparse Mixture-of-Experts Transformer Model. With 25B active parameters and 218B total parameters, it has 128 experts, out of which 8 are active per token, and a single shared expert is applied to all tokens. The attention layers interleave sliding-window attention layers with Rotational Positional Embeddings and global attention layers without positional embeddings in a 3:1 ratio, as first introduced in Command A. The sparse MoE layer is trained in a fully dropless manner and uses a token-choice router. We use additive-bias-based load balancing to encourage balanced token load across all experts, and swap out the softmax router activation function with a normalized sigmoid over the topk expert logits per token. 
+**Model Architecture**: Command A+ is a decoder-only Sparse Mixture-of-Experts Transformer Model. With 25B active parameters and 218B total parameters, it has 128 experts, out of which 8 are active per token, and a single shared expert is applied to all tokens. The attention layers interleave sliding-window attention layers with Rotational Positional Embeddings and global attention layers without positional embeddings in a 3:1 ratio, as first introduced in Command A. The sparse MoE layer is trained in a fully dropless manner and uses a token-choice router. We use additive-bias-based load balancing to encourage balanced token load across all experts, and swap out the softmax router activation function with a normalized sigmoid over the topk expert logits per token.
 
 **Languages covered:** The model has been trained on 48 languages: English, Arabic, Bulgarian, Bengali, Catalan, Czech, Danish, German, Greek, Spanish, Estonian, Persian, Finnish, Filipino, French, Irish, Hebrew, Hindi, Croatian, Hungarian, Indonesian, Icelandic, Italian, Japanese, Korean, Lithuanian, Latvian, Malay, Maltese, Dutch, Norwegian, Punjabi, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Serbian, Swedish, Tamil, Telugu, Thai, Turkish, Ukrainian, Urdu, Vietnamese, Chinese.
 
@@ -295,5 +294,3 @@ For errors or additional questions about details in this model card, contact \[[
 **Try it now:**
 
 You can try Command A+ in the [playground](https://dashboard.cohere.com/playground/chat?model=command-a-plus-05-2026). You can also use it in our dedicated [Hugging Face Space](https://huggingface.co/spaces/CohereLabs/command-a-plus-05-2026).
-
-