Text Generation
Transformers
Safetensors
step3p5
conversational
custom_code
fp8
WinstonDeng commited on
Commit
0071250
·
verified ·
1 Parent(s): d417eb0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -5
README.md CHANGED
@@ -9,7 +9,7 @@ base_model:
9
  <div align="center">
10
 
11
  <div align="center" style="display: flex; justify-content: center; align-items: center;">
12
- <img src="https://huggingface.co/stepfun-ai/Step-3.5-Flash/resolve/main/figures/stepfun.svg" width="25" style="margin-right: 10px;"/>
13
  <h1 style="margin: 0; border-bottom: none;">Step-3.5-Flash</h1>
14
  </div>
15
 
@@ -38,7 +38,7 @@ base_model:
38
 
39
  Step 3.5 Flash delivers performance parity with leading closed-source systems while remaining open and efficient.
40
 
41
- ![]
42
 
43
  Performance of Step 3.5 Flash measured across **Reasoning**, **Coding**, and **Agency**. Open-source models (left) are sorted by their total parameter count, while top-tier proprietary models are shown on the right. xbench-DeepSearch scores are sourced from [official publications](https://xbench.org/agi/aisearch) for consistency. The shadowed bars represent the enhanced performance of Step 3.5 Flash using [Parallel Thinking](https://arxiv.org/pdf/2601.05593).
44
 
@@ -219,6 +219,8 @@ pip install "sglang[all] @ git+https://github.com/sgl-project/sglang.git"
219
  ```
220
 
221
  2. Launch the server.
 
 
222
  - For bf16 model
223
  SGLANG_ENABLE_SPEC_V2=1
224
  python3 -m sglang.launch_server \
@@ -657,7 +659,7 @@ python3 -m sglang.launch_server \
657
  - When using `vLLM` and `SGLang`, thinking mode is enabled by default when sending requests.
658
  - Both support tool calling. Please use OpenAI-style tool description format for calls.
659
 
660
- ## Citation
661
 
662
  If you find our work useful in your research, please consider citing the following paper:
663
 
@@ -671,8 +673,8 @@ If you find our work useful in your research, please consider citing the followi
671
  primaryClass={cs.CL},
672
  url={https://arxiv.org/abs/xxxxx},
673
  }
674
- ```
675
 
676
  ## 📄 License
677
 
678
- This project is open-sourced under the [Apache 2.0 License](https://www.google.com/search?q=LICENSE).
 
9
  <div align="center">
10
 
11
  <div align="center" style="display: flex; justify-content: center; align-items: center;">
12
+ <img src="stepfun.svg" width="25" style="margin-right: 10px;"/>
13
  <h1 style="margin: 0; border-bottom: none;">Step-3.5-Flash</h1>
14
  </div>
15
 
 
38
 
39
  Step 3.5 Flash delivers performance parity with leading closed-source systems while remaining open and efficient.
40
 
41
+ ![](step-bar-chart.png)
42
 
43
  Performance of Step 3.5 Flash measured across **Reasoning**, **Coding**, and **Agency**. Open-source models (left) are sorted by their total parameter count, while top-tier proprietary models are shown on the right. xbench-DeepSearch scores are sourced from [official publications](https://xbench.org/agi/aisearch) for consistency. The shadowed bars represent the enhanced performance of Step 3.5 Flash using [Parallel Thinking](https://arxiv.org/pdf/2601.05593).
44
 
 
219
  ```
220
 
221
  2. Launch the server.
222
+
223
+ ```bash
224
  - For bf16 model
225
  SGLANG_ENABLE_SPEC_V2=1
226
  python3 -m sglang.launch_server \
 
659
  - When using `vLLM` and `SGLang`, thinking mode is enabled by default when sending requests.
660
  - Both support tool calling. Please use OpenAI-style tool description format for calls.
661
 
662
+ <!-- ## Citation
663
 
664
  If you find our work useful in your research, please consider citing the following paper:
665
 
 
673
  primaryClass={cs.CL},
674
  url={https://arxiv.org/abs/xxxxx},
675
  }
676
+ ``` -->
677
 
678
  ## 📄 License
679
 
680
+ This project is open-sourced under the [Apache 2.0 License](https://www.google.com/search?q=LICENSE).