Sign up today and get $10 in free AI credits to test out the API

neutrino logo

neutrino AI

Neutrino Inference Engine

Increase quality, decrease costs, reduce latency on LLM inference

Powered by Intelligent Model Routing

Try our Inference Engine

Creating optimal inference for your Coding Copilots

Leverage a collection of LLMs

We constantly evaluate the latest models to make sure you get the best intelligence

Intelligently route to the best model

Model routing directs each query to the best-suited model for each task

Remove reliance on a single model provider

Automatic load-balancing and outage handling to ensure maximum uptime and reliability

Improve your response quality with task-specific engines

Some models are great at coding, some are better at reasoning, some excel at text summarization.

Maximize performance by using the best model selection for your AI tasks and intelligently routing to the optimal model for each query.

Optimize your AI costs

While GPT4 is great at reasoning, it is often unnecessary for the majority of tasks. Querying the optimal model can result in 40 - 95% lower AI costs per month.
GPT-4
Claude 3 Sonnet
Mistral 7B Instruct

Minimize response latency

Improve your user experience by minimizing response latency by only calling the larger, slower models when they're actually needed.

Powered by intelligent model routing

How Neutrino routes each query to the best-suited model.
clusters

Data Collection and Use-case Analysis

We understand different LLM use-cases by collecting and clustering prompts, and generating responses for different models.
balance

Model Benchmarking & Performance Evaluation

We evaluate and asses model performance across a variety of use-cases, and benchmark top-performing models.
intelligence

Training the Model Router

We train a collection of prediction models to identify in real-time, which models are likely to generate high-quality responses for a given query.

Benchmarks

MT-Bench

Neutrino Router is roughly 98.9% as accurate as GPT4 while being 46.87% cheaper in MT-Bench Evaluations.

OpenAI Evals

Evaluated on more than a dozen random test datasets from OpenAI Eval Repo
Neutrino Router is 96.9% as accurate as GPT-4 but 50% cheaper.
In some tests Neutrino Router had savings up to 95%
By comparison GPT-3.5-Turbo is only 60.2% as accurate as GPT-4

Start with a familiar API

We designed our API to be compatible with OpenAI's to make the transition as seamless as possible.
Python
LangChain
1from openai import OpenAI
2
3client = OpenAI(
4    base_url="https://router.neutrinoapp.com/api/engines",
5    api_key="<Neutrino-API-key>"
6)
7
8client.chat.completions.create(
9    model="code" # options: code or chat,
10    messages = [{"role": "user", "content": "Hello, how are you?"}],
11)
12

Pricing

Choose between our standard plan offering pre-built engines for common AI use cases or our enterprise plan for custom solutions.

Standard

15%

of AI spend

  • Unlimited API requests
  • Router fine-tuning
  • Access to all models
get started

Enterprise

Custom

Talk to sales

  • Fine-tuned models
  • Private deployment
  • Priority feature requests
contact sales

Checkout our inference prices for specific models <<MODEL TOKEN PRICING>>

Sign up to access our platform

get started