Google Gemma 4 Claims Top Rankings on Japanese Swallow Leaderboard v2

Google Gemma

May 12, 2026 · Updated May 20, 2026

Google confirmed that its Gemma 4 open-weight model achieved high-ranking results on the Swallow Leaderboard v2, a rigorous Japanese language benchmark. This validation establishes the model as a leading choice for developers building regional applications that require frontier-level reasoning in Japanese.

Google confirmed that its Gemma 4 open-weight model (where the trained parameters are publicly released) achieved high-ranking results on the Swallow Leaderboard v2. This benchmark (a standardized test for evaluating model quality) assesses Japanese language capabilities to ensure global models perform effectively across regional linguistic and reasoning tasks.

This shift highlights the closing gap between global open-weight models and specialized regional systems. While previous Gemma 4 manual resolution controls focused on vision and Gemma 4 drafter models improved speed, these results prove linguistic versatility. It positions Google's open models as viable alternatives to proprietary systems in major markets.

You can now use Gemma 4 for Japanese-language applications requiring high-performance reasoning. The model weights are available for self-hosting or via the Google AI Studio API. This follows the recent integration of Gemini CLI's local routing which allows Gemma models to handle intent analysis on-device.

Google Gemma

@googlegemmaMay 11

Gemma 4 is a highly-capable model in Japanese! Amazing to see great results for it on the latest Swallow Leaderboard v2 results. ✨ Excited to see what’s next in the open model space! https://t.co/aK4raCKIeh

110801

View on X

Still wondering? A few quick answers below.

The Swallow Leaderboard v2 is a specialized benchmark developed by the Tokyo Institute of Technology to evaluate the Japanese language capabilities of large language models. It uses a variety of tests, including math, coding, and general reasoning, to rank how effectively models handle linguistic nuances and complex tasks specifically within the Japanese language context.

Google Gemma 4 is an open-weight model, meaning its trained parameters are publicly released for developers to download and run. While it is built on Google proprietary Gemini technology, the open-weight nature allows for local deployment and customization, making it a flexible choice for researchers and developers who prefer self-hosting over cloud-only APIs.

Google Gemma 4 is a highly capable model for Japanese language tasks, as evidenced by its high-ranking results on the Swallow Leaderboard v2. It demonstrates strong performance across diverse benchmarks such as JHumanEval for coding and MATH-100 for mathematics, establishing it as a leading open-weight option for applications requiring deep Japanese linguistic understanding.

Developers can access Google Gemma 4 through the Google AI Studio API for cloud-based integration or by downloading the open weights for local inference. Because it is a lightweight model family, it is designed to be efficient enough for deployment on various hardware configurations, including edge devices, while maintaining frontier-level performance for Japanese and English tasks.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from Google →

Keep reading

Arena Ranks Google Gemma 4 as Top Open Vision Model

Google's Gemma-4-31b and Gemma-4-26b-a4b have entered the Vision Arena leaderboard as the #2 and #4 ranked open models. These releases shift the price-performance frontier by delivering vision reasoning capabilities that rival proprietary systems at a fraction of the cost.

GoogleApr 27

Google Launches Gemma 4 to Bring Frontier Reasoning to Local Devices

Google released Gemma 4, a new family of open models built on the same architecture as Gemini 3 and licensed under Apache 2.0. These models deliver high-performance reasoning and native multimodal capabilities directly on consumer hardware, enabling private, offline agentic workflows. This shift allows developers to build sophisticated AI applications that run entirely on-device without sacrificing intelligence.

Google Gemma 4 31B Leads Open-Weight Models in Negotiation Benchmark

Google GemmaMay 28

Google Gemma 4 31B Leads Open-Weight Models in Negotiation Benchmark

Google's Gemma 4 31B ranked as the top-performing open-weight model on TERMS-Bench, a new evaluation for AI agents conducting economic negotiations. The benchmark uses a verifiable environment instead of LLM grading to measure an agent's ability to maximize profit while following strict financial constraints.

Artificial Analysis Benchmarks Google's Gemma 4 12B Transcription at 8.8% WER

Artificial AnalysisJun 6

Artificial Analysis Benchmarks Google's Gemma 4 12B Transcription at 8.8% WER

Artificial Analysis benchmarked Google DeepMind's new open-weight Gemma 4 12B model for transcription, reporting an 8.8% Word Error Rate (WER). This places the model behind specialized open-weight transcription solutions, but it is available for local deployment alongside Google's new Eloquent dictation app.

What is the Swallow Leaderboard v2?

Is Google Gemma 4 an open-source model?

How does Google Gemma 4 perform in the Japanese language?

How can developers access and use Google Gemma 4?

Keep reading

Arena Ranks Google Gemma 4 as Top Open Vision Model

Arena Ranks Google Gemma 4 as Top Open Vision Model

Google Launches Gemma 4 to Bring Frontier Reasoning to Local Devices

Google Launches Gemma 4 to Bring Frontier Reasoning to Local Devices

Google Gemma 4 31B Leads Open-Weight Models in Negotiation Benchmark

Google Gemma 4 31B Leads Open-Weight Models in Negotiation Benchmark

Artificial Analysis Benchmarks Google's Gemma 4 12B Transcription at 8.8% WER

Artificial Analysis Benchmarks Google's Gemma 4 12B Transcription at 8.8% WER

Keep reading

Arena Ranks Google Gemma 4 as Top Open Vision Model

Arena Ranks Google Gemma 4 as Top Open Vision Model

Google Launches Gemma 4 to Bring Frontier Reasoning to Local Devices

Google Launches Gemma 4 to Bring Frontier Reasoning to Local Devices

Google Gemma 4 31B Leads Open-Weight Models in Negotiation Benchmark

Google Gemma 4 31B Leads Open-Weight Models in Negotiation Benchmark

Artificial Analysis Benchmarks Google's Gemma 4 12B Transcription at 8.8% WER

Artificial Analysis Benchmarks Google's Gemma 4 12B Transcription at 8.8% WER