Instructions to use ruv/ruvltra with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ruv/ruvltra with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="ruv/ruvltra",
	filename="ruvltra-claude-code-0.5b-q4_k_m.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 ruv/ruvltra with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf ruv/ruvltra:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf ruv/ruvltra:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf ruv/ruvltra:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf ruv/ruvltra:Q4_K_M

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 ruv/ruvltra:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf ruv/ruvltra:Q4_K_M

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 ruv/ruvltra:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ruv/ruvltra:Q4_K_M

Use Docker

docker model run hf.co/ruv/ruvltra:Q4_K_M

LM Studio
Jan

vLLM

How to use ruv/ruvltra with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/ruv/ruvltra:Q4_K_M

Ollama
How to use ruv/ruvltra with Ollama:
```
ollama run hf.co/ruv/ruvltra:Q4_K_M
```

Unsloth Studio new

How to use ruv/ruvltra 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 ruv/ruvltra 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 ruv/ruvltra to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for ruv/ruvltra to start chatting

Docker Model Runner
How to use ruv/ruvltra with Docker Model Runner:
```
docker model run hf.co/ruv/ruvltra:Q4_K_M
```

Lemonade

How to use ruv/ruvltra with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ruv/ruvltra:Q4_K_M

Run and chat with the model

lemonade run user.ruvltra-Q4_K_M

List all available models

lemonade list

ruv commited on Jan 21

Commit

e7de02d

verified ·

1 Parent(s): c316a38

data: Add v2.5 performance optimization stats

Browse files

Files changed (1) hide show

training/v2.5-performance-stats.json +67 -0

training/v2.5-performance-stats.json ADDED Viewed

	@@ -0,0 +1,67 @@

+{
+  "version": "2.5",
+  "release_name": "Performance Optimized Edition",
+  "release_date": "2026-01-21T10:46:53.928251",
+  "optimizations": {
+    "hnsw_index": {
+      "description": "Hierarchical Navigable Small World graphs",
+      "improvement": "10x faster search at 10k entries"
+    },
+    "lru_cache": {
+      "description": "O(1) LRU cache using Rust lru crate",
+      "lookup_time_ns": 23.5
+    },
+    "zero_copy": {
+      "description": "Arc<str> string interning",
+      "improvement": "100-1000x cache improvement"
+    },
+    "batch_simd": {
+      "description": "AVX2/NEON vectorization",
+      "improvement": "4x throughput"
+    },
+    "memory_pools": {
+      "description": "Arena allocation",
+      "improvement": "50% fewer allocations"
+    }
+  },
+  "benchmarks": {
+    "query_decomposition_ns": 340,
+    "cache_lookup_ns": 23.5,
+    "memory_search_10k_ms": 0.4,
+    "pattern_retrieval_us": 25,
+    "routing_accuracy_hybrid": 1.0,
+    "routing_accuracy_embedding_only": 0.45
+  },
+  "models": {
+    "claude_code_0.5b": {
+      "file": "ruvltra-claude-code-0.5b-q4_k_m.gguf",
+      "size_mb": 398,
+      "purpose": "Agent routing",
+      "context_length": 32768
+    },
+    "small_0.5b": {
+      "file": "ruvltra-small-0.5b-q4_k_m.gguf",
+      "size_mb": 400,
+      "purpose": "General embeddings",
+      "context_length": 32768
+    },
+    "medium_3b": {
+      "file": "ruvltra-medium-3b-q4_k_m.gguf",
+      "size_mb": 2048,
+      "purpose": "Full LLM inference",
+      "context_length": 262144
+    }
+  },
+  "performance_targets": {
+    "flash_attention_speedup": "2.49x-7.47x",
+    "hnsw_search_speedup": "150x-12500x",
+    "memory_reduction": "50-75%",
+    "mcp_response_ms": 100,
+    "sona_adaptation_ms": 0.05
+  },
+  "training_data": {
+    "labeled_examples": 381,
+    "contrastive_pairs": 793,
+    "agent_types": 60
+  }
+}