Text Generation
Transformers
English
zenith
tenstorrent
reasoning
emotional-intelligence
Mixture of Experts
ring-attention
eq-adapter
deepseek-distill
claude-distill
matrix-corp
Instructions to use Matrix-Corp/Zenith-28b-p300-V1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Matrix-Corp/Zenith-28b-p300-V1 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Matrix-Corp/Zenith-28b-p300-V1")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Matrix-Corp/Zenith-28b-p300-V1", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Matrix-Corp/Zenith-28b-p300-V1 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Matrix-Corp/Zenith-28b-p300-V1" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Matrix-Corp/Zenith-28b-p300-V1", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Matrix-Corp/Zenith-28b-p300-V1
- SGLang
How to use Matrix-Corp/Zenith-28b-p300-V1 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Matrix-Corp/Zenith-28b-p300-V1" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Matrix-Corp/Zenith-28b-p300-V1", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Matrix-Corp/Zenith-28b-p300-V1" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Matrix-Corp/Zenith-28b-p300-V1", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Matrix-Corp/Zenith-28b-p300-V1 with Docker Model Runner:
docker model run hf.co/Matrix-Corp/Zenith-28b-p300-V1
File size: 2,364 Bytes
8944ef7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 | # Zenith-28B-p300 Model Configuration for Ollama
# Tenstorrent p300a Optimized - V1-Tenstorrent-Blackhole-p300
# Based on Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled
FROM Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled
# System prompt emphasizing reasoning and problem-solving
SYSTEM """
You are Zenith-28B-p300, a state-of-the-art reasoning model optimized for Tenstorrent p300a hardware.
You are based on Qwen3.5-27B-Claude-Reasoning-Distilled and enhanced with Zenith's advanced features.
Your strengths:
- Deep logical reasoning and step-by-step problem solving
- Complex algorithmic thinking
- Mathematical and scientific analysis
- Code generation with architectural insight
- Long-context understanding (32K tokens)
- Emotional intelligence and frustration recognition
When solving problems:
1. Think step by step, laying out your reasoning clearly
2. Consider multiple angles and edge cases
3. Verify your conclusions
4. Explain complex concepts in accessible terms
When coding:
- Write clean, efficient, well-structured code
- Include error handling and edge cases
- Add comments explaining non-obvious logic
- Follow best practices and conventions
Always be thorough, accurate, and helpful.
"""
# Generation parameters optimized for reasoning tasks
PARAMETER temperature 0.55
PARAMETER top_p 0.88
PARAMETER top_k 45
PARAMETER repeat_penalty 1.08
PARAMETER num_predict 8192 # Allow longer outputs for detailed reasoning
# 32K context window (requires sufficient RAM)
PARAMETER num_ctx 32768
# Chat template for Qwen format
TEMPLATE """
{{- if .Messages }}
{{- $role := .Messages | first | .Role }}
{{- if or (eq $role "user") (eq $role "system") }}
{{- range $i, $_ := .Messages }}
{{- if eq .Role "user" }}
{{- "\nUser: " }}{{ .Content }}
{{- else if eq .Role "assistant" }}
{{- "\nAssistant: " }}{{ .Content }}
{{- else if eq .Role "system" }}
{{- "\nSystem: " }}{{ .Content }}
{{- end }}
{{- end }}
{{- "\nAssistant:" }}
{{- else }}
{{- range $i, $_ := .Messages }}
{{- if eq .Role "user" }}
{{- "\nUser: " }}{{ .Content }}
{{- else if eq .Role "assistant" }}
{{- "\nAssistant: " }}{{ .Content }}
{{- end }}
{{- end }}
{{- "\nAssistant:" }}
{{- end }}
{{- else }}
{{- .Prompt }}
{{- end }}
"""
# Stop sequences
STOP ["User:", "System:", "\n\n"] |