Lokalog

🌟 What’s new?

We’ve added more visibility to how your Custom AI Profiles are configured and performing. You can now see exactly how many examples are available for AI to use as context when generating translations:

• Tags: Shows the number of tagged keys available for a profile and AI to use as context
• Reviewed translations: Displays the number of reviewed translations from selected projects available for AI reference

🎯 Why it matters to you?

One of the most common questions we’ve heard from users is how many examples are “enough” for Custom AI Profiles to work effectively. With this update, you now have a clear view of the available context so you can make sure your profiles are set up for the best translation quality, consistency, and brand alignment.

🌟 What’s new?

We’ve upgraded to Claude 4.5 Sonnet, Anthropic’s latest and most capable model for translations and AI scoring. This ensures you always have access to the newest and most stable language technology available for your localization workflows.

🎯 Why it matters to you?

Based on our internal testing, the model is comparable to Claude 4 Sonnet. So far we see acceptance rates that match but do not surpass the top achievements of the previous version. The upgrade maintains the high-quality translation output you expect while providing access to Anthropic's latest model improvements in instruction following and reasoning capabilities.

🌟 What’s new?

We’ve added a new report to the Tasks dashboard called “Detailed task data” that is available for customers on the Enterprise plan and above.
This report gives you a granular view of task performance, with a row for every task × language × contributor combination. It’s designed for advanced analytics and export use cases, making it easier to dig deeper into task efficiency, processed word counts, Translation Memory (TM) usage, and contributor performance across all your projects.

What’s included

• Full task details (ID, title, project, type, source/target languages, creation/due/completion dates)
• Assignment information (created by, completed by, closed by)
• Time-to-complete metrics
• Translation Memory match breakdowns (0–49%, 50–74%, 75–84%, 85–94%, 95–99%, 100%)

🎯 Why it matters to you?

This report helps you get a complete picture of your team’s localization efficiency and performance — making it easier to identify bottlenecks, measure productivity, and optimize TM usage.

🌟 What’s new?

We’re excited to share that Lokalise was part of the official launch of Figma Buzz, Figma’s new tool for managing and localizing content directly within design workflows. Our Lokalise Plugin for Figma Buzz is available from day one, allowing you to translate and manage your design content seamlessly inside Figma Buzz.

🎯 Why it matters to you?

You can now connect design and localization teams earlier in the process, ensuring brand-consistent translations right where content is created. Launching alongside Figma Buzz puts Lokalise at the heart of Figma’s growing ecosystem — empowering you to manage translations more efficiently from design to delivery.

🌟 What’s new?

Custom AI Profiles, powered by Advanced Retrieval augmented generation (RAG) let you improve AI translation output by providing examples from your own past translations and project context. Profiles allow you to:

• Specify source and target languages, pulling relevant TM entries for those pairs.
• Apply a TM date filter to limit examples to a specific timeframe.
• Activate profiles per project to control where they apply.

Currently, three retrieval sources are supported:

• Translation Memory (TM) – historical translations
• Tags – filter examples by project or string-level tags (e.g., product area, tone)
• Reviewed translations – prioritize quality-approved examples"

🎯 Why it matters to you?

You’ll see higher translation quality, consistency, and brand alignment, as AI mirrors your team’s preferred wording and style. This also reduces manual edits, boosting efficiency and increasing acceptance rates to 90%+.

👉 Interested in trying this powerful feature early? Join our Beta program →

🌟 What’s new?

We’ve added an option in the Workflow editor for Automatic Translation Tasks to automatically save translations performed by AI or MT directly to your Translation Memory.

🎯 Why it matters to you?

Your translations are now automatically captured in your TM, reducing manual steps and ensuring that all AI/MT translations contribute to your reusable content, making future translations faster and more consistent.

🌟 What’s new?

Our Python SDK for the Lokalise API has been upgraded to version 4 with a completely reworked typing system. 

The main change: the entire typing system has been rebuilt. This means the SDK now provides complete and accurate type annotations for all methods, models, and responses. Editors and tools like Pyright, PyCharm, VS Code, and mypy can now understand the SDK structure properly.

🎯 Why it matters to you?

Developers now get type hinting directly in their code editors, making it easier to write, understand, and maintain code when working with the SDK. This leads to faster development, fewer errors, and a smoother integration experience.

🌟 What’s new?

Workflow steps that require assignees — for translation or review — now automatically assign contributors based on their language and role. This matches the logic used in task creation and eliminates the need for manual assignment.

Note: Currently, you need to navigate to the specific translation or review task step for assignees to populate. This will be fixed in a future update.

🎯 Why it matters to you?

You’ll save time and effort, especially when working with many target languages, as contributors are assigned automatically. This makes workflow setup faster, reduces errors, and keeps your localization process moving smoothly.

🌟 What’s new?

Our GitHub Actions for file exchange have been upgraded to version 4, introducing a lighter, more efficient package that no longer relies on a CLI tool. The update also includes various performance and reliability improvements.

🎯 Why it matters to you?

You’ll benefit from faster, more stable file exchange between Lokalise and GitHub, with improved error handling. Most customers won’t be affected by breaking changes, but you can review the full release notes here:

• Pull Action v4
• Push Action v4

🌟 What’s new?

We’ve upgraded to GPT-5, the latest model from OpenAI. This update improves translation accuracy and consistency, especially for languages with less training data, and strengthens AI scoring for more reliable results.

🎯 Why it matters to you?

You’ll notice more stable, high-quality outputs in your AI translations with no extra setup required. GPT-5’s stronger instruction-following means translations better reflect your source content and style preferences, helping your team work faster with fewer edits.