---

Tessa-T1-3B-GGUF

Original Model

Tesslate/Tessa-T1-3B

Run with LlamaEdge

• LlamaEdge version: v0.16.14 and above

• Prompt template

  • Prompt type: chatml

    <|im_start|>system
    {system_message}<|im_end|>
    <|im_start|>user
    {prompt}<|im_end|>
    <|im_start|>assistant
    

  • Prompt type: chatml-think

    <|im_start|>system
    {system_message}<|im_end|>
    <|im_start|>user
    {prompt}<|im_end|>
    <|im_start|>assistant
    <|im_start|>think
    

• Context size: 32000

• Run as LlamaEdge service

  wasmedge --dir .:. --nn-preload default:GGML:AUTO:Tessa-T1-3B-Q5_K_M.gguf \
    llama-api-server.wasm \
    --model-name Tessa-T1-3B \
    --prompt-template chatml \
    --ctx-size 32000
  
  # Think mode
  wasmedge --dir .:. --nn-preload default:GGML:AUTO:Tessa-T1-3B-Q5_K_M.gguf \
    llama-api-server.wasm \
    --model-name Tessa-T1-3B \
    --prompt-template chatml-think \
    --ctx-size 32000
  

• Run as LlamaEdge command app

  wasmedge --dir .:. --nn-preload default:GGML:AUTO:Tessa-T1-3B-Q5_K_M.gguf \
    llama-chat.wasm \
    --prompt-template chatml \
    --ctx-size 32000
  
  # Think mode
  wasmedge --dir .:. --nn-preload default:GGML:AUTO:Tessa-T1-3B-Q5_K_M.gguf \
    llama-chat.wasm \
    --prompt-template chatml-think \
    --ctx-size 32000
  

Quantized GGUF Models

Name	Quant method	Bits	Size	Use case
Tessa-T1-3B-Q2_K.gguf	Q2_K	2	1.27 GB	smallest, significant quality loss - not recommended for most purposes
Tessa-T1-3B-Q3_K_L.gguf	Q3_K_L	3	1.71 GB	small, substantial quality loss
Tessa-T1-3B-Q3_K_M.gguf	Q3_K_M	3	1.59 GB	very small, high quality loss
Tessa-T1-3B-Q3_K_S.gguf	Q3_K_S	3	1.45 GB	very small, high quality loss
Tessa-T1-3B-Q4_0.gguf	Q4_0	4	1.82 GB	legacy; small, very high quality loss - prefer using Q3_K_M
Tessa-T1-3B-Q4_K_M.gguf	Q4_K_M	4	1.93 GB	medium, balanced quality - recommended
Tessa-T1-3B-Q4_K_S.gguf	Q4_K_S	4	1.83 GB	small, greater quality loss
Tessa-T1-3B-Q5_0.gguf	Q5_0	5	2.17 GB	legacy; medium, balanced quality - prefer using Q4_K_M
Tessa-T1-3B-Q5_K_M.gguf	Q5_K_M	5	2.22 GB	large, very low quality loss - recommended
Tessa-T1-3B-Q5_K_S.gguf	Q5_K_S	5	2.17 GB	large, low quality loss - recommended
Tessa-T1-3B-Q6_K.gguf	Q6_K	6	2.54 GB	very large, extremely low quality loss
Tessa-T1-3B-Q8_0.gguf	Q8_0	8	3.29 GB	very large, extremely low quality loss - not recommended
Tessa-T1-3B-f16.gguf	f16	16	6.18 GB

Quantized with llama.cpp b4944