Repository files navigation

🚀 2-10x accuracy improvements on reasoning tasks with zero training

🤗 HuggingFace Space •📓 Colab Demo •💬 Discussions

OptiLLM is an OpenAI API-compatible optimizing inference proxy that implements 20+ state-of-the-art techniques to dramatically improve LLM accuracy and performance on reasoning tasks - without requiring any model training or fine-tuning.

It is possible to beat the frontier models using these techniques across diverse tasks by doing additional compute at inference time. A good example of how to combine such techniques together is theCePO approach from Cerebras.

• 🎯 Instant Improvements: 2-10x better accuracy on math, coding, and logical reasoning
• 🔌 Drop-in Replacement: Works with any OpenAI-compatible API endpoint
• 🧠 20+ Optimization Techniques: From simple best-of-N to advanced MCTS and planning
• 📦 Zero Training Required: Just proxy your existing API calls through OptiLLM
• ⚡ Production Ready: Used in production by companies and researchers worldwide
• 🌍 Multi-Provider: Supports OpenAI, Anthropic, Google, Cerebras, and 100+ models via LiteLLM

Get powerful reasoning improvements in 3 simple steps:

# 1. Install OptiLLMpip install optillm# 2. Start the serverexport OPENAI_API_KEY="your-key-here"optillm# 3. Use with any OpenAI client - just change the model name!

fromopenaiimportOpenAIclient=OpenAI(base_url="http://localhost:8000/v1")# Add 'moa-' prefix for Mixture of Agents optimizationresponse=client.chat.completions.create(model="moa-gpt-4o-mini",# This gives you GPT-4o performance from GPT-4o-mini!messages=[{"role":"user","content":"Solve: If 2x + 3 = 7, what is x?"}])

Before OptiLLM: "x = 1" ❌
After OptiLLM: "Let me work through this step by step: 2x + 3 = 7, so 2x = 4, therefore x = 2" ✅

OptiLLM delivers measurable improvements across diverse benchmarks:

Full benchmark resultsbelow ⬇️

pip install optillmoptillm2024-10-22 07:45:05,612 - INFO - Loaded plugin: privacy2024-10-22 07:45:06,293 - INFO - Loaded plugin: memory2024-10-22 07:45:06,293 - INFO - Starting server with approach: auto

docker pull ghcr.io/codelion/optillm:latestdocker run -p 8000:8000 ghcr.io/codelion/optillm:latest2024-10-22 07:45:05,612 - INFO - Loaded plugin: privacy2024-10-22 07:45:06,293 - INFO - Loaded plugin: memory2024-10-22 07:45:06,293 - INFO - Starting server with approach: auto

Available Docker image variants:

• Full image (latest): Includes all dependencies for local inference and plugins
• Proxy-only (latest-proxy): Lightweight image without local inference capabilities
• Offline (latest-offline): Self-contained image with pre-downloaded models (spaCy) for fully offline operation

# Proxy-only (smallest)docker pull ghcr.io/codelion/optillm:latest-proxy# Offline (largest, includes pre-downloaded models)docker pull ghcr.io/codelion/optillm:latest-offline

Clone the repository withgit and usepip install to setup the dependencies.

git clone https://github.com/codelion/optillm.gitcd optillmpython3 -m venv .venvsource .venv/bin/activatepip install -r requirements.txt

OptILLM supports SSL certificate verification configuration for working with self-signed certificates or corporate proxies.

Disable SSL verification (development only):

# Command lineoptillm --no-ssl-verify# Environment variableexport OPTILLM_SSL_VERIFY=falseoptillm

Use custom CA certificate:

# Command lineoptillm --ssl-cert-path /path/to/ca-bundle.crt# Environment variableexport OPTILLM_SSL_CERT_PATH=/path/to/ca-bundle.crtoptillm

⚠️Security Note: Disabling SSL verification is insecure and should only be used in development. For production environments with custom CAs, use--ssl-cert-path instead. SeeSSL_CONFIGURATION.md for details.

We support all major LLM providers and models for inference. You need to set the correct environment variable and the proxy will pick the corresponding client.

You can then run the optillm proxy as follows.

python optillm.py2024-09-06 07:57:14,191 - INFO - Starting server with approach: auto2024-09-06 07:57:14,191 - INFO - Server configuration: {'approach':'auto','mcts_simulations': 2,'mcts_exploration': 0.2,'mcts_depth': 1,'best_of_n': 3,'model':'gpt-4o-mini','rstar_max_depth': 3,'rstar_num_rollouts': 5,'rstar_c': 1.4,'base_url':''}* Serving Flask app'optillm'* Debug mode: off2024-09-06 07:57:14,212 - INFO - WARNING: This is a development server. Do not use itin a production deployment. Use a production WSGI server instead.* Running on all addresses (0.0.0.0)* Running on http://127.0.0.1:8000* Running on http://192.168.10.48:80002024-09-06 07:57:14,212 - INFO - Press CTRL+C to quit

Once the proxy is running, you can use it as a drop in replacement for an OpenAI client by setting thebase_url ashttp://localhost:8000/v1.

importosfromopenaiimportOpenAIOPENAI_KEY=os.environ.get("OPENAI_API_KEY")OPENAI_BASE_URL="http://localhost:8000/v1"client=OpenAI(api_key=OPENAI_KEY,base_url=OPENAI_BASE_URL)response=client.chat.completions.create(model="moa-gpt-4o",messages=[    {"role":"user","content":"Write a Python program to build an RL model to recite text from any position that the user provides, using only numpy."    }  ],temperature=0.2)print(response)

The code above applies to both OpenAI and Azure OpenAI, just remember to populate theOPENAI_API_KEY env variable with the proper key.There are multiple ways to control the optimization techniques, they are applied in the follow order of preference:

• You can control the technique you use for optimization by prepending the slug to the model name{slug}-model-name. E.g. in the above code we are usingmoa or mixture of agents as the optimization approach. In the proxy logs you will see the following showing themoa is been used with the base model asgpt-4o-mini.

2024-09-06 08:35:32,597 - INFO - Using approach moa, with gpt-4o-mini2024-09-06 08:35:35,358 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions"HTTP/1.1 200 OK"2024-09-06 08:35:39,553 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions"HTTP/1.1 200 OK"2024-09-06 08:35:44,795 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions"HTTP/1.1 200 OK"2024-09-06 08:35:44,797 - INFO - 127.0.0.1 - - [06/Sep/2024 08:35:44]"POST /v1/chat/completions HTTP/1.1" 200 -

• Or, you can pass the slug in theoptillm_approach field in theextra_body.

response = client.chat.completions.create(  model="gpt-4o-mini",  messages=[{"role":"user","content":"" }],  temperature=0.2,  extra_body={"optillm_approach":"bon|moa|mcts"})

• Or, you can just mention the approach in either yoursystem oruser prompt, within<optillm_approach> </optillm_approach> tags.

response = client.chat.completions.create(  model="gpt-4o-mini",  messages=[{"role":"user","content":"<optillm_approach>re2</optillm_approach> How many r's are there in strawberry?" }],  temperature=0.2)

Please note that the convention described above works only when the optillm server has been started with inference approach set toauto. Otherwise, themodel attribute in the client request must be set with the model name only.

We now support all LLM providers (by wrapping around theLiteLLM sdk). E.g. you can use the Gemini Flash model withmoa by setting passing the api key in the environment variableos.environ['GEMINI_API_KEY'] and then calling the modelmoa-gemini/gemini-1.5-flash-002. In the output you will then see that LiteLLM is being used to call the base model.

9:43:21 - LiteLLM:INFO: utils.py:2952 -LiteLLMcompletion() model= gemini-1.5-flash-002; provider = gemini2024-09-29 19:43:21,011 - INFO -LiteLLMcompletion() model= gemini-1.5-flash-002; provider = gemini2024-09-29 19:43:21,481 - INFO - HTTP Request: POST https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash-002:generateContent?key=[redacted]"HTTP/1.1 200 OK"19:43:21 - LiteLLM:INFO: utils.py:988 - Wrapper: Completed Call, calling success_handler2024-09-29 19:43:21,483 - INFO - Wrapper: Completed Call, calling success_handler19:43:21 - LiteLLM:INFO: utils.py:2952 -LiteLLMcompletion() model= gemini-1.5-flash-002; provider = gemini

The following sequence diagram illustrates how the request and responses go through optillm.

In the diagram:

• A is an existing tool (likeoobabooga), framework (likepatchwork)or your own code where you want to use the results from optillm. You can use it directly using any OpenAI client sdk.
• B is the optillm service (running directly or in a docker container) that will send requests to thebase_url.
• C is any service providing an OpenAI API compatible chat completions endpoint.

We support loading any HuggingFace model or LoRA directly in optillm. To use the built-in inference server set theOPTILLM_API_KEY to any value (e.g.export OPTILLM_API_KEY="optillm")and then use the same in your OpenAI client. You can pass any HuggingFace model in model field. If it is a private model make sure you set theHF_TOKEN environment variablewith your HuggingFace key. We also support adding any number of LoRAs on top of the model by using the+ separator.

E.g. The following code loads the base modelmeta-llama/Llama-3.2-1B-Instruct and then adds two LoRAs on top -patched-codes/Llama-3.2-1B-FixVulns andpatched-codes/Llama-3.2-1B-FastApply.You can specify which LoRA to use using theactive_adapter param inextra_body field of OpenAI SDK client. By default we will load the last specified adapter.

OPENAI_BASE_URL="http://localhost:8000/v1"OPENAI_KEY="optillm"response=client.chat.completions.create(model="meta-llama/Llama-3.2-1B-Instruct+patched-codes/Llama-3.2-1B-FastApply+patched-codes/Llama-3.2-1B-FixVulns",messages=messages,temperature=0.2,logprobs=True,top_logprobs=3,extra_body={"active_adapter":"patched-codes/Llama-3.2-1B-FastApply"},)

You can also use the alternate decoding techniques likecot_decoding andentropy_decoding directly with the local inference server.

response=client.chat.completions.create(model="meta-llama/Llama-3.2-1B-Instruct",messages=messages,temperature=0.2,extra_body={"decoding":"cot_decoding",# or "entropy_decoding"# CoT specific params"k":10,"aggregate_paths":True,# OR Entropy specific params"top_k":27,"min_p":0.03,    })

• Set theOPENAI_API_KEY env variable to a placeholder value
  • e.g.export OPENAI_API_KEY="sk-no-key"
• Run./llama-server -c 4096 -m path_to_model to start the server with the specified model and a context length of 4096 tokens
• Runpython3 optillm.py --base_url base_url to start the proxy
  • e.g. for llama.cpp, runpython3 optillm.py --base_url http://localhost:8080/v1

The Model Context Protocol (MCP) plugin enables OptiLLM to connect with MCP servers, bringing external tools, resources, and prompts into the context of language models. This allows for powerful integrations with filesystem access, database queries, API connections, and more.

OptiLLM supports bothlocal andremote MCP servers through multiple transport methods:

• stdio: Local servers (traditional)
• SSE: Remote servers via Server-Sent Events
• WebSocket: Remote servers via WebSocket connections

TheModel Context Protocol (MCP) is an open protocol standard that allows LLMs to securely access tools and data sources through a standardized interface. MCP servers can provide:

• Tools: Callable functions that perform actions (like writing files, querying databases, etc.)
• Resources: Data sources for providing context (like file contents)
• Prompts: Reusable prompt templates for specific use cases

Note on Backwards Compatibility: Existing MCP configurations will continue to work unchanged. Thetransport field defaults to "stdio" when not specified, maintaining full backwards compatibility with existing setups.

1. Create a configuration file at~/.optillm/mcp_config.json with the following structure:

Local Server (stdio) - Traditional Method:

{"mcpServers": {"filesystem": {"transport":"stdio","command":"npx","args": ["-y","@modelcontextprotocol/server-filesystem","/path/to/allowed/directory1","/path/to/allowed/directory2"      ],"env": {},"description":"Local filesystem access"    }  },"log_level":"INFO"}

Legacy Format (still works):

{"mcpServers": {"filesystem": {"command":"npx","args": ["-y","@modelcontextprotocol/server-filesystem","/path/to/directory"],"env": {}    }  }}

Remote Server (SSE) - New Feature:

{"mcpServers": {"github": {"transport":"sse","url":"https://api.githubcopilot.com/mcp","headers": {"Authorization":"Bearer ${GITHUB_TOKEN}","Accept":"text/event-stream"      },"timeout":30.0,"sse_read_timeout":300.0,"description":"GitHub MCP server for repository access"    }  },"log_level":"INFO"}

Remote Server (WebSocket) - New Feature:

{"mcpServers": {"remote-ws": {"transport":"websocket","url":"wss://api.example.com/mcp","description":"Remote WebSocket MCP server"    }  },"log_level":"INFO"}

Mixed Configuration (Local + Remote):

{"mcpServers": {"filesystem": {"transport":"stdio","command":"npx","args": ["-y","@modelcontextprotocol/server-filesystem","/home/user/docs"],"description":"Local filesystem access"    },"github": {"transport":"sse","url":"https://api.githubcopilot.com/mcp","headers": {"Authorization":"Bearer ${GITHUB_TOKEN}"      },"description":"GitHub MCP server"    },"remote-api": {"transport":"websocket","url":"wss://api.company.com/mcp","description":"Company internal MCP server"    }  },"log_level":"INFO"}

Common Parameters:

• Server name: A unique identifier for the server (e.g., "filesystem", "github")
• transport: Transport method - "stdio" (default), "sse", or "websocket"
• description (optional): Description of the server's functionality
• timeout (optional): Connection timeout in seconds (default: 5.0)

stdio Transport (Local Servers):

• command: The executable to run the server
• args: Command-line arguments for the server
• env: Environment variables for the server process

sse Transport (Server-Sent Events):

• url: The SSE endpoint URL
• headers (optional): HTTP headers for authentication
• sse_read_timeout (optional): SSE read timeout in seconds (default: 300.0)

websocket Transport (WebSocket):

• url: The WebSocket endpoint URL

Environment Variable Expansion:Headers and other string values support environment variable expansion using${VARIABLE_NAME} syntax. This is especially useful for API keys:

{"headers": {"Authorization":"Bearer ${GITHUB_TOKEN}","X-API-Key":"${MY_API_KEY}"  }}

OptiLLM supports both local and remote MCP servers:

You can use any of theofficial MCP servers or third-party servers that run as local processes:

• Filesystem:@modelcontextprotocol/server-filesystem - File operations
• Git:mcp-server-git - Git repository operations
• SQLite:@modelcontextprotocol/server-sqlite - SQLite database access
• Brave Search:@modelcontextprotocol/server-brave-search - Web search capabilities

Remote servers provide centralized access without requiring local installation:

• GitHub MCP Server:https://api.githubcopilot.com/mcp - Repository management, issue tracking, and code analysis
• Third-party servers: Any MCP server that supports SSE or WebSocket protocols

{"mcpServers": {"filesystem": {"transport":"stdio","command":"npx","args": ["-y","@modelcontextprotocol/server-filesystem","/home/user/documents"],"description":"Local file system access"    },"search": {"transport":"stdio","command":"npx","args": ["-y","@modelcontextprotocol/server-brave-search"],"env": {"BRAVE_API_KEY":"your-api-key-here"      },"description":"Web search capabilities"    },"github": {"transport":"sse","url":"https://api.githubcopilot.com/mcp","headers": {"Authorization":"Bearer ${GITHUB_TOKEN}","Accept":"text/event-stream"      },"description":"GitHub repository and issue management"    }  },"log_level":"INFO"}

Once configured, the MCP plugin will automatically:

1. Connect to all configured MCP servers
2. Discover available tools, resources, and prompts
3. Make these capabilities available to the language model
4. Handle tool calls and resource requests

The plugin enhances the system prompt with MCP capabilities so the model knows which tools are available. When the model decides to use a tool, the plugin:

1. Executes the tool with the provided arguments
2. Returns the results to the model
3. Allows the model to incorporate the results into its response

Here are some examples of queries that will engage MCP tools:

Local Server Examples:

• "List all the Python files in my documents directory" (Filesystem)
• "What are the recent commits in my Git repository?" (Git)
• "Search for the latest information about renewable energy" (Search)
• "Query my database for all users who registered this month" (Database)

Remote Server Examples:

• "Show me the open issues in my GitHub repository" (GitHub MCP)
• "Create a new branch for the feature I'm working on" (GitHub MCP)
• "What are the most recent pull requests that need review?" (GitHub MCP)
• "Get the file contents from my remote repository" (GitHub MCP)

The MCP plugin logs detailed information to:

~/.optillm/logs/mcp_plugin.log

Check this log file for connection issues, tool execution errors, and other diagnostic information.

Local Server Issues (stdio transport):

1. Command not found: Make sure the server executable is available in your PATH, or use an absolute path in the configuration.

2. Access denied: For filesystem operations, ensure the paths specified in the configuration are accessible to the process.

Remote Server Issues (SSE/WebSocket transport):

3. Connection timeout: Remote servers may take longer to connect. Increase thetimeout value in your configuration.

4. Authentication failed: Verify your API keys and tokens are correct. For GitHub MCP server, ensure yourGITHUB_TOKEN environment variable is set with appropriate permissions.

5. Network errors: Check your internet connection and verify the server URL is accessible.

6. Environment variable not found: If using${VARIABLE_NAME} syntax, ensure the environment variables are set before starting OptILLM.

General Issues:

7. Method not found: Some servers don't implement all MCP capabilities (tools, resources, prompts). Verify which capabilities the server supports.

8. Transport not supported: Ensure you're using a supported transport: "stdio", "sse", or "websocket".

Example: Testing GitHub MCP Connection

To test if your GitHub MCP server configuration is working:

1. Set your GitHub token:export GITHUB_TOKEN="your-github-token"
2. Start OptILLM and check the logs at~/.optillm/logs/mcp_plugin.log
3. Look for connection success messages and discovered capabilities

optillm supports various command-line arguments for configuration. When using Docker, these can also be set as environment variables prefixed withOPTILLM_.

optillm can optionally be built and run using Docker and the providedDockerfile.

1. Make sure you have Docker and Docker Compose installed on your system.

2. Either update the environment variables in the docker-compose.yaml file or create a.env file in the project root directory and add any environment variables you want to set. For example, to set the OpenAI API key, add the following line to the.env file:

   OPENAI_API_KEY=your_openai_api_key_here

3. Run the following command to start optillm:

   docker compose up -d

   This will build the Docker image if it doesn't exist and start the optillm service.

4. optillm will be available athttp://localhost:8000.

When using Docker, you can set these parameters as environment variables. For example, to set the approach and model, you would use:

OPTILLM_APPROACH=mctsOPTILLM_MODEL=gpt-4

To secure the optillm proxy with an API key, set theOPTILLM_API_KEY environment variable:

OPTILLM_API_KEY=your_secret_api_key

When the API key is set, clients must include it in their requests using theAuthorization header:

Authorization: Bearer your_secret_api_key

Model: google/gemini-2.5-flash-lite-preview-09-2025 via OpenRouterConfiguration: 3 agents, 2-pass verification, thinking tags disabled for proofs

¹ Performance numbers reported by LongBench v2 authors, except for LongCePO and Llama-4-Maverick results.

² Numbers in parentheses for LongCePO indicate accuracy of majority voting from 5 runs.

¹ Numbers in parentheses for LongCePO indicate accuracy of majority voting from 5 runs.

Since optillm is a drop-in replacement for OpenAI API you can easily integrate it with existing tools and frameworks using the OpenAI client. We used optillm withpatchwork which is an open-source framework that automates development gruntwork like PR reviews, bug fixing, security patching using workflowscalled patchflows. We saw huge performance gains across all the supported patchflows as shown below when using the mixture of agents approach (moa).

OptiLLM includes a comprehensive test suite to ensure reliability and compatibility.

The main test suite can be run from the project root:

# Test all approaches with default test casespython tests/test.py# Test specific approachespython tests/test.py --approaches moa bon mcts# Run a single testpython tests/test.py --single-test"Simple Math Problem"

Additional tests are available in thetests/ directory:

# Run all tests (requires pytest)./tests/run_tests.sh# Run specific test modulespytest tests/test_plugins.py -vpytest tests/test_api_compatibility.py -v

All tests are automatically run on pull requests via GitHub Actions. The workflow tests:

• Multiple Python versions (3.10, 3.11, 3.12)
• Unit tests for plugins and core functionality
• API compatibility tests
• Integration tests with various approaches

Seetests/README.md for more details on the test structure and how to write new tests.

We ❤️ contributions! OptiLLM is built by the community, for the community.

• 🐛Found a bug?Open an issue
• 💡Have an idea?Start a discussion
• 🔧Want to code? Check outgood first issues

git clone https://github.com/codelion/optillm.gitcd optillmpython -m venv .venvsource .venv/bin/activate# or `.venv\Scripts\activate` on Windowspip install -r requirements.txtpip install -r tests/requirements.txt# Run testspython -m pytest tests/

• Eliciting Fine-Tuned Transformer Capabilities via Inference-Time Techniques
• AutoThink: efficient inference for reasoning LLMs -Implementation
• Deep Think with Confidence: Confidence-guided reasoning and inference-time scaling -Implementation
• Self-Discover: Large Language Models Self-Compose Reasoning Structures -Implementation
• CePO: Empowering Llama with Reasoning using Test-Time Compute -Implementation
• LongCePO: Empowering LLMs to efficiently leverage infinite context -Implementation
• Chain of Code: Reasoning with a Language Model-Augmented Code Emulator -Inspired the implementation of coc plugin
• Entropy Based Sampling and Parallel CoT Decoding -Implementation
• Fact, Fetch, and Reason: A Unified Evaluation of Retrieval-Augmented Generation -Evaluation script
• Writing in the Margins: Better Inference Pattern for Long Context Retrieval -Inspired the implementation of the memory plugin
• Chain-of-Thought Reasoning Without Prompting -Implementation
• Re-Reading Improves Reasoning in Large Language Models -Implementation
• In-Context Principle Learning from Mistakes -Implementation
• Planning In Natural Language Improves LLM Search For Code Generation -Implementation
• Self-Consistency Improves Chain of Thought Reasoning in Language Models -Implementation
• Mutual Reasoning Makes Smaller LLMs Stronger Problem-Solvers -Implementation
• Mixture-of-Agents Enhances Large Language Model Capabilities -Inspired the implementation of moa
• Prover-Verifier Games improve legibility of LLM outputs -Implementation
• Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning -Inspired the implementation of mcts
• Unsupervised Evaluation of Code LLMs with Round-Trip Correctness -Inspired the implementation of rto
• Patched MOA: optimizing inference for diverse software development tasks -Implementation
• Patched RTC: evaluating LLMs for diverse software development tasks -Implementation
• AIMO-2 Winning Solution: Building State-of-the-Art Mathematical Reasoning Models with OpenMathReasoning dataset -Implementation
• Test-Time Diffusion Deep Researcher (TTD-DR): Think More, Research More, Answer Better! -Implementation

If you use this library in your research, please cite:

@software{optillm,title ={OptiLLM: Optimizing inference proxy for LLMs},author ={Asankhaya Sharma},year ={2024},publisher ={GitHub},url ={https://github.com/codelion/optillm}}

Ready to optimize your LLMs? Install OptiLLM and see the difference! 🚀

⭐Star us on GitHub if you find OptiLLM useful!