DiffusionGemma 26B A4B IT is an open-weights multimodal generative model developed by Google DeepMind that processes text, image, and video inputs to produce text output via discrete diffusion. Built on the Gemma 4 26B A4B Mixture-of-Experts (MoE) architecture with 25.2B total parameters and 3.8B active parameters, the model employs an encoder-decoder design with bidirectional attention that generates tokens in parallel 256-token blocks, enabling high-speed generation exceeding 1,100 tokens per second at low batch sizes on NVIDIA Hopper H100 (FP8). DiffusionGemma 26B A4B IT supports a 256K token context window, configurable thinking (reasoning) mode, native function calling, and multilingual inference across 35+ languages.
This model is ready for commercial or non-commercial use.
This model is not owned or developed by NVIDIA. This model has been developed and built to a third-party's requirements for this application and use case; see link to Non-NVIDIA DiffusionGemma 26B A4B IT Model Card
GOVERNING TERMS: The trial service is governed by the NVIDIA API Trial Terms of Service. Use of the model is governed by the NVIDIA Open Model Agreement. Additional Information: Apache 2.0. Gemma Terms of Use and Gemma Prohibited Use Policy.
Global
Use Case: DiffusionGemma 26B A4B IT is designed for developers, researchers, and enterprises requiring high-speed multimodal text generation. Supported use cases include conversational AI and chatbots, text summarization, code generation and step-by-step reasoning, image and document understanding (OCR, chart comprehension, PDF parsing, screen and UI parsing), video content analysis, agentic workflows with native function calling, and multilingual NLP tasks across 35+ languages.
Build.NVIDIA.com: 06/2026 via link
Huggingface: 06/10/2026 via link
References:
Architecture Type: Transformer
Network Architecture: Mixture-of-Experts
Total Parameters: 25.2B
Active Parameters: 3.8B
Vocabulary Size: 262,144
Base Model: Gemma 4 26B A4B
Input Types: Text, Image, Video
Input Formats: String, Red, Green, Blue (RGB), Video (MP4)
Input Parameters: One-Dimensional (1D), Two-Dimensional (2D), Three-Dimensional (3D)
Other Input Properties: Images support variable aspect ratios and resolutions via a configurable visual token budget (70, 140, 280, 560, or 1120 tokens per image); place image content before text for optimal multimodal performance. Videos are processed as frame sequences up to 60 seconds at 1 frame per second.
Input Context Length (ISL): 262,144
Output Types: Text
Output Format: String
Output Parameters: One-Dimensional (1D)
Other Output Properties: Text is generated in parallel 256-token canvas blocks via diffusion sampling with adaptive stopping; supports native function calling and structured JSON output formatting.
Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.
Runtime Engines:
Supported Hardware:
Operating Systems: Linux
The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.
DiffusionGemma 26B A4B IT v1.0 (June 2026)
Data Modality: Text, Image
Text Training Data Size: Undisclosed
Image Training Data Size: Undisclosed
Training Data Collection: Hybrid: Automated, Human
Training Labeling: Hybrid: Automated, Human
Training Properties: Large-scale, diverse pre-training corpus spanning web documents, code, mathematics, and images across 140+ languages with a data collection cutoff of January 2025. Preprocessing includes CSAM filtering, sensitive personal information removal, and content quality filtering aligned with Google's AI policies.
Testing Data Collection: Undisclosed
Testing Labeling: Undisclosed
Testing Properties: Undisclosed
Evaluation Benchmark Score: DiffusionGemma 26B A4B IT achieves strong performance across text reasoning, code, vision, and long-context benchmarks, including 77.6% on MMLU Pro, 73.2% on GPQA Diamond, and 70.5% on MATH-Vision. All results are for instruction-tuned model variants evaluated with the recommended Entropy Bound (EB) sampler.
| Benchmark | DiffusionGemma 26B A4B | Gemma 4 26B A4B |
|---|---|---|
| Text & Reasoning | ||
| MMLU Pro | 77.6% | 82.6% |
| AIME 2026 (no tools) | 69.1% | 88.3% |
| LiveCodeBench v6 | 69.1% | 77.1% |
| Codeforces ELO | 1429 | 1718 |
| GPQA Diamond | 73.2% | 82.3% |
| Tau2 (average over 3) | 56.2% | 68.2% |
| HLE (no tools) | 11.0% | 8.7% |
| HLE (with search) | 11.9% | 17.2% |
| BigBench Extra Hard | 47.6% | 64.8% |
| MMMLU | 81.5% | 86.3% |
| Vision | ||
| MMMU Pro | 54.3% | 73.8% |
| OmniDocBench 1.5 (avg edit distance, lower is better) | 0.319 | 0.149 |
| MATH-Vision | 70.5% | 82.4% |
| MedXPertQA MM | 49.0% | 58.1% |
| Long Context | ||
| MRCR v2 8 needle 128k (average) | 32.0% | 44.1% |
Evaluations conducted on instruction-tuned variants using the recommended Entropy-Bounded Denoising with Adaptive Stopping (EB) sampler.
Evaluation Data Collection: Automated
Evaluation Labeling: Automated
Evaluation Properties: Standard benchmark evaluation across 15 tasks covering text reasoning, code generation, multilingual understanding, vision, and long-context retrieval. All evaluations were performed on instruction-tuned model variants using the Entropy-Bounded Denoising with Adaptive Stopping (EB) sampler (max 48 denoising steps, temperature linear decay 0.8 → 0.4, entropy bound 0.1, adaptive stopping entropy threshold 0.005). All testing was intentionally conducted without safety filters to evaluate the model's raw capabilities and baseline behaviors.
Acceleration Engine: vLLM
Test Hardware: H100
DiffusionGemma 26B A4B IT generates text via block-autoregressive discrete diffusion. An autoregressive encoder processes the input prompt and caches it via KV cache; a decoder then applies bidirectional attention over a 256-token generation canvas, iteratively denoising the full block in parallel. Recommended sampling settings:
Simpler tasks (e.g., structured code output) require fewer denoising steps, enabling dynamic throughput scaling by task complexity.
Enable step-by-step reasoning by including <|think|> at the start of the system prompt. Reasoning output is wrapped in <|channel>thought\n[reasoning]<channel|> before the final answer. If thinking is disabled, empty thought tags are still emitted. In multi-turn conversations, do not include prior thinking content in conversation history — include only the final response.
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. Developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
Please make sure you have proper rights and permissions for all input image and video content; if image or video includes people, personal health information, or intellectual property, the image or video generated will not blur or maintain proportions of image subjects included.
Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here