> This page is for version v5.0 (default).
> For other versions, use one of these documentation indexes:
> - v5.0 (default): https://humanloop.com/docs/v5/llms.txt
> - v4.0: https://humanloop.com/docs/v4/llms.txt

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://humanloop.com/docs/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://humanloop.com/docs/_mcp/server.

# February

## OpenAI GPT 4.5 preview model

*February 27th, 2025*

We've added OpenAI GPT 4.5 preview model to Humanloop the day it was released!

![Screenshot showing GPT 4.5 Preview](/docs/_fern-img/bbbc68bbaf5d00c072cb757139386e86a7d67b0d36ca25bb0893ef0e182d0628.webp)

GPT-4.5 delivers more natural responses thanks to its broader knowledge base, enhanced ability to follow user intent, and improved emotional intelligence.
It excels in tasks like writing enhancement, programming, and problem-solving. To find out more, read the [OpenAI blog post](https://openai.com/index/introducing-gpt-4-5/).

## Claude 3.7 Sonnet model

*February 24th, 2025*

We've added Anthropic Claude 3.7 to Humanloop the day it was released!

![Screenshot showing Claude 3.7 Sonnet](/docs/_fern-img/5dd310c3d21043ce4e29f0c415122436f9d4bf8b21c09bb47d9ad49b278f92f5.webp)

Claude 3.7 Sonnet can produce near-instant responses or extended, step-by-step thinking.
Additionally, it shows particularly strong improvements in coding and front-end web development. To find out more, read the [Anthropic blog post](https://www.anthropic.com/news/claude-3-7-sonnet).

## Improved Log Display

*February 20th, 2025*

To significantly improve readability in your use cases with strict formatting requirements, you can now choose to view your log contents in various formats. Formats such as JSON, YAML, Markdown, and more will be automatically detected and made available for manual selection.

![Screenshot showing the improved Log Display](/docs/_fern-img/337623dd05ddbc0bb922da97b1da44f31b4671645d3df5cd77ca53d880cf1983.webp)

To try it, simply open up a Log from the Logs or Evals tab and hover over its output content.

## Jinja Support

*February 17th, 2025*

We've added support for Jinja in the Editor. [Jinja](https://jinja.palletsprojects.com/en/stable/) is a powerful templating engine that allows you to define variables, conditionals and other logic. You can now leverage this directly in the text of your Prompt Templates.

To try it out, navigate to the Editor tab and select Jinja as the templating language in the new dropdown in the **Template** section:

![Screenshot showing Jinja in Editor](/docs/_fern-img/238d4010607738a4fedcf3e8899337d10a20cc048eb31e877662215a51ae5496.webp)

We've also added a new API endpoint `prompts/{id}/populate` that takes the `inputs` required by your template and returns the Prompt version along with the populated template.

## Improved Stats Page in Evaluations

*February 14th, 2025*

We've revamped our Stats page in Evaluations, which shows a summary of the evaluation results across runs and compares them to a control run.

![Screenshot showing the improved Stats page in Evaluations](/docs/_fern-img/aa55810e59916a2e00cb1b014dcf3242f26c300437aca78ac3ccc9bff9c6ffa7.webp)

Some highlights:

* Collapsible sections to view the most important top-line metric for each evaluator at a glance
* New and improved bar charts for categorial and boolean evaluators
* Clear percentage-based comparisons against the control run for each metric
* Click through from metric values and comparisons to a filtered Review page to drill down into datapoints

## Gemini 2.0 Flash is now available

*February 12th, 2025*

We've added support for the latest Google Gemini 2.0 Flash model on Humanloop.

![Screenshot showing reasoning effort in the Editor](/docs/_fern-img/b23383db3abd343e70412ca225ed09353da3dba166bdfe166d2a6eeeef5f445d.webp)

Gemini Flash 2.0 is optimal for high-volume, high-frequency tasks. It is highly capable of multimodal reasoning across vast amounts of information, with a context window of 1 million tokens.

## Reasoning effort in Editor

*February 11th, 2025*

You can now specify the reasoning effort for o3-mini and o1 models in Editor.

![Screenshot showing reasoning effort in the Editor](/docs/_fern-img/62764d568ec59cdd41bc6c70956d36ff83b0ae987d6f023aee1255b6a0b04a49.webp)

The reasoning effort parameter gives the model guidance on how many reasoning tokens it should generate before creating a response to the prompt. For more complex tasks, you may want to increase the reasoning effort to allow the model to 'think' longer before responding.

## Structured Outputs in Editor

*February 3rd, 2025*

You can now specify the model’s response format directly in Editor. This is especially useful when you need the model to respond in a defined structure, such as JSON.

![Screenshot showing structured outputs in the Editor](/docs/_fern-img/beb667c247ac8b1eb0cac125c1e56db384e314992411eb18b82711ed43cbf09a.webp)

To try it out:

> 1. Select a **Prompt**.
> 2. Open the **Editor** tab.
> 3. In the **Response Format** dropdown at the bottom, select **JSON Schema**.

You can then define your JSON schema manually or use our AI-powered JSON schema generator by clicking the Generate button above the schema editor.