Today, Vantage is launching a free, open-source website that provides a centralized view of LLM model specifications, pricing, benchmarks, and hosting information at vantage.sh/models. The site is publicly available and supported as a community resource by Vantage. If you’re familiar with https://instances.vantage.sh/, the site formerly known as https://ec2instances.info/, this is the LLM equivalent of that.

Information about LLMs today is often fragmented. Some LLM comparison sites focus primarily on pricing, while others focus on benchmarks or model capabilities. It can be hard to get all the information you want in a single place. Existing sites also tend to offer limited flexibility for customizing comparisons or queries.

To help with this, Vantage has launched vantage.sh/models, which combines model specifications, pricing, benchmarks, and hosting information in a single, intuitive interface that is open-source and continuously updated, so users have a centralized place to access and compare all model information instead of relying on multiple separate sources.

The home page includes filtering and customizable querying across LLM vendors, reasoning tier, benchmark performance, pricing, token limits, release dates, hosting options, and more, so users can compare models side by side. Users can easily choose which columns to display, creating comparison views tailored to their needs. For more advanced use cases, the columns displayed on the site are generated by editable SQL queries that can be modified to support custom comparison views. This allows flexible querying, such as comparing models from one vendor in a specific region against the average pricing of another vendor.

Each model has its own dedicated detail page, which includes a pricing calculator, tokenizer, and more in-depth specifications for that specific model. All model data is pulled programmatically from model provider APIs and public benchmark sources to ensure it remains current.

Head to vantage.sh/models to explore and compare leading LLMs in one place! To contribute or suggest improvements, please go to GitHub.

Frequently Asked Questions

1. What is being launched today?

Vantage is launching vantage.sh/models, a free and open-source website that provides a centralized view of LLM model specifications, pricing, benchmarks, and hosting information.

2. Who is the customer?

vantage.sh/models is publicly available to anyone evaluating or building with LLMs, including engineers, AI teams, FinOps practitioners, and analysts.

3. How much does this cost?

There is no cost to use vantage.sh/models. It is publicly available and does not require a Vantage account.

4. How is Vantage supporting this?

vantage.sh/models is maintained by Vantage as part of its broader commitment to improving transparency in cloud and AI infrastructure costs. The site is open-source and aligned with Vantage’s mission to help teams better understand and optimize their infrastructure spend.

5. How can I contribute or make a suggestion?

The project is open-source on GitHub, where users can submit issues, suggest additional models or benchmarks, and contribute improvements.

6. What other features will be added to vantage.sh/models?

Future enhancements may include additional benchmark integrations, expanded provider coverage, historical pricing data, model comparison views, and API access for programmatic retrieval of model data.

7. How often is data updated on vantage.sh/models?

Data on vantage.sh/models is updated daily. Model specifications, pricing information, and benchmark data are refreshed daily from provider APIs and maintained public data sources to ensure the site remains current.

8. What models are included on this site?

vantage.sh/models includes a wide range of LLMs across major providers. This includes models such as GPT-series models, Claude models, Gemini models, Llama models, DeepSeek models, Mistral models, and others. The site focuses on models that are available through public APIs or managed model platforms and that provide publicly documented pricing or specifications. New models are added on an ongoing basis as they become available.

9. What vendors are included in vantage.sh/models?

The site includes models from major model developers and hosting providers. Current vendors include OpenAI, Anthropic, Google, Meta, DeepSeek, Mistral, IBM, Nvidia, and Qwen. Models are also tracked across hosting platforms such as OpenAI API, AWS Bedrock, Google Cloud Vertex AI, IBM watsonx.ai, Anthropic API, and Mistral API. Additional vendors and hosting platforms will be added over time.

10. What model metrics are collected for benchmarking?

vantage.sh/models aggregates benchmark results from publicly available sources. Current metrics include Humanity’s Last Exam scores, SWE-Bench Resolved percentages, and SkateBench scores where available. The site also includes non-benchmark specifications such as context window size, token limits, release date, training cutoff date, reasoning support, and self-hosting availability. Benchmark coverage varies by model depending on publicly available data.

11. Are APIs available to programmatically access this data?

At launch, vantage.sh/models does not provide a public API. However, the site is open-source, and the underlying data and queries are publicly accessible. You can also pull down a copy of the SQLite3 database with /data.db. API access and structured data exports may be added in the future.

13. What additional data can be added/displayed?

The site uses a query-driven data model that allows new columns and comparison views to be defined using SQL queries. Queries run dynamically for each model and can combine data from multiple tables, including model specifications, vendor availability, regional pricing, tokenizer information, and benchmark results. This makes it possible to create domain-specific views, such as percentage cost differences between cached and uncached tokens or vendor-specific pricing comparisons.