Instructions to use dcostenco/prism-coder-32b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dcostenco/prism-coder-32b with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="dcostenco/prism-coder-32b",
	filename="prism-coder-32b-q4km.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use dcostenco/prism-coder-32b with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf dcostenco/prism-coder-32b
# Run inference directly in the terminal:
llama-cli -hf dcostenco/prism-coder-32b

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf dcostenco/prism-coder-32b
# Run inference directly in the terminal:
llama-cli -hf dcostenco/prism-coder-32b

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf dcostenco/prism-coder-32b
# Run inference directly in the terminal:
./llama-cli -hf dcostenco/prism-coder-32b

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf dcostenco/prism-coder-32b
# Run inference directly in the terminal:
./build/bin/llama-cli -hf dcostenco/prism-coder-32b

Use Docker

docker model run hf.co/dcostenco/prism-coder-32b

LM Studio
Jan

vLLM

How to use dcostenco/prism-coder-32b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dcostenco/prism-coder-32b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dcostenco/prism-coder-32b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dcostenco/prism-coder-32b

Ollama
How to use dcostenco/prism-coder-32b with Ollama:
```
ollama run hf.co/dcostenco/prism-coder-32b
```

Unsloth Studio new

How to use dcostenco/prism-coder-32b with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for dcostenco/prism-coder-32b to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for dcostenco/prism-coder-32b to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for dcostenco/prism-coder-32b to start chatting

Pi new

How to use dcostenco/prism-coder-32b with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf dcostenco/prism-coder-32b

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "dcostenco/prism-coder-32b"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use dcostenco/prism-coder-32b with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf dcostenco/prism-coder-32b

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default dcostenco/prism-coder-32b

Run Hermes

hermes

Docker Model Runner
How to use dcostenco/prism-coder-32b with Docker Model Runner:
```
docker model run hf.co/dcostenco/prism-coder-32b
```

Lemonade

How to use dcostenco/prism-coder-32b with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull dcostenco/prism-coder-32b

Run and chat with the model

lemonade run user.prism-coder-32b-{{QUANT_TAG}}

List all available models

lemonade list

dcostenco commited on 16 days ago

Commit

46e84ae

verified ·

1 Parent(s): 70fdf61

docs: v27 prompt — 14B hits 98% (ties Claude Opus), labeled-category routing breakthrough

Browse files

Files changed (1) hide show

README.md +31 -57

README.md CHANGED Viewed

@@ -14,73 +14,47 @@ tags:
 - prism-coder
 ---
-# prism-coder:32b (v19)
-LoRA fine-tune of **Qwen/QwQ-32B** for offline MCP tool routing — Synalux Copilot's "reasoning tier", used when 14B can't decide or when a complex multi-step prompt arrives.
-## Test results — Prism routing 100-case eval (May 14 2026)
-100 prompts (seed=2027), v26 system prompt + nothink template.
-| Category | Current | Previous (v19 old prompt) | Δ |
-|---|---|---|---|
-| **Overall** | **97.3% ± 0.6%** | 93.7% | **+3.6** |
-| session_load_context | 100% | 100% | = |
-| session_save_ledger | 100% | 100% | = |
-| session_search_memory | 100% | 100% | = |
-| session_save_handoff | 100% | 100% | = |
-| session_compact_ledger | 100% | 100% | = |
-| brave_web_search | 100% | 100% | = |
-| knowledge_search | 100% | 100% | = |
-| AAC plain-text | **100%** | 79% | **+21** |
-| translate plain-text | 83% | 83% | = |
-| plain text (pred/irrel) | 100% | 67% | +33 |
-| no-tool refusal | 100% | 100% | = |
-| info / lookup | 100% | 100% | = |
-| edge (multi-step) | 80% | 82% | -2 |
-| **avg latency** | **2.7s** | 2.3s | +0.4s |
-| **invented tools** | 0 | 0 | = |
-**Key improvement (May 14 2026)**: system prompt v26 eliminates Q4_K_M quantization artifacts where "plain text" was misread as a tool name. AAC routing jumped from 79% to 100% — critical for the life-critical AAC use case.
-**97.3% puts this model within 2 points of Claude Sonnet 4 (99%) on the same eval**, while running fully offline on a Mac.
-Only 2 misroutes in 100 cases: "Convert 'good morning' to Japanese" → brave_web_search (edge case), and a multi-step ledger query.
-**What this benchmark measures**: routing precision against the *exact* 7-tool Prism Coder taxonomy. **Not** a general-capability score. Methodology + runner: [github.com/dcostenco/prism-coder/tree/main/tests/benchmarks/prism-routing-100](https://github.com/dcostenco/prism-coder/tree/main/tests/benchmarks/prism-routing-100).
-## Notes on QwQ-32B base
-QwQ-32B natively emits `<think>...</think>` blocks. The Ollama Modelfile uses a `nothink` template (pre-closes the `<think>` block) to skip reasoning and go straight to the tool call.
-## Training recipe (v19)
-- **Base**: Qwen/QwQ-32B
-- **LoRA**: r=32, α=64, dropout 0.05, all 7 target modules
-- **Corpus**: ~14K-row composite (Phase 1 general + Phase 2 agentic + Phase 3 multi-turn XL)
-- **Hardware**: RunPod A100 80GB / RTX 6000 Ada
-- **Quantization**: published as Q4_K_M GGUF (~19 GB) and merged HF safetensors
 ## Usage
-### Ollama (recommended)
 ```bash
 ollama pull dcostenco/prism-coder:32b
-ollama run dcostenco/prism-coder:32b "Save handoff for prism-coder — deployment complete"
 ```
-### System prompt
-Use the [v26 routing prompt](https://github.com/dcostenco/prism-coder/blob/main/tests/benchmarks/prism-routing-100/benchmark.py#L47) with the `nothink` template.
 ## Hardware requirements
-- **Mac**: M2 Ultra+ with ≥48 GB unified memory (Q4_K_M = 19 GB + activations)
-- **Linux + NVIDIA**: A100 40GB+, H100, B200, or 2× RTX 4090
-- **Inference speed**: ~2–3 s per 200-token response
 - **Loaded VRAM**: ~22 GB
 ## License
-Inherits QwQ-32B's Apache-2.0 license. The LoRA delta is also Apache-2.0.

 - prism-coder
 ---
+# prism-coder:32b (v19) — 97.3% routing accuracy
+LoRA fine-tune of **Qwen/QwQ-32B** for offline MCP tool routing — Synalux Copilot's "reasoning tier."
+## Test results — Prism routing 100-case eval (May 15 2026, 3-seed mean)
+| Category | Score |
+|---|---|
+| **Overall** | **97.3% ± 0.6%** |
+| session_load_context | 100% |
+| session_save_ledger | 100% |
+| session_search_memory | 100% |
+| session_save_handoff | 100% |
+| session_compact_ledger | 100% |
+| brave_web_search | 100% |
+| knowledge_search | 100% |
+| AAC plain-text | 85% |
+| translate plain-text | 83% |
+| plain text (pred/irrel) | 100% |
+| no-tool refusal | 100% |
+| info / lookup | 100% |
+| edge (multi-step) | 100% |
+| **avg latency** | **2.4s** |
+| **invented tools** | 0 |
+All 7 MCP tools route correctly 100% of the time. Remaining misses are in plain-text categories (AAC, translate) where the model occasionally over-routes.
+Uses `nothink` template + [v27 system prompt](https://github.com/dcostenco/prism-coder/blob/main/tests/benchmarks/prism-routing-100/benchmark.py#L47) with labeled category headers.
 ## Usage
 ```bash
 ollama pull dcostenco/prism-coder:32b
 ```
 ## Hardware requirements
+- **Mac**: M2 Ultra+ with ≥48 GB unified memory
+- **Linux + NVIDIA**: A100 40GB+, H100, B200
 - **Loaded VRAM**: ~22 GB
 ## License
+Apache-2.0 (inherits from QwQ-32B).