Skip to content

Models

GenAI4Science portal is currently optimized for enabling multiple models, so that users can try as much as possible. Various larger model sizes of the DeepSeek, Gemma, GLM, Llama, Mistral, Muse, OpenAI and Qwen model families are available, including versions optimized for coding and for visual recognition.

Since not every model can fit in the memory of the GPU cards at the same time, it happens that you have to wait a bit when changing models.

Each backend instance can handle limited number of requests per model at the same time. If this limit has been reached, then the request is waiting in the backend queue. When this queue is full, the following error message is displayed:

Ollama: 503, message='Service Unavailable'

After a short wait, you should press the "Regenerate" button below the message to try again. If this message still occurs, it is advisable to choose another model.

To protect the UI from overload, we have implemented a rate limit.

"Rate limit exceeded. Please try again later."

After a short wait, you should press the "Regenerate" button.

Available language models

Our models are regularly updated, and the list below may not reflect the currently available models. For the most up-to-date information on available models and their brief descriptions, please log in to access the UI. If you require a specific model that is not currently installed, please contact us.

You can find information about context sizes in the UI model’s brief description section, or HERE.

Model tags

Each model in the UI is labeled with one or more tags indicating its main capabilities:

  • Reasoning – The model produces step-by-step "thinking" (chain-of-thought) before its final answer, improving performance on math, logic, and complex multi-step problems.
  • Vision – The model is multimodal: it accepts images as input in addition to text.
  • Code – The model is optimized for programming tasks: code generation, debugging, and agentic coding workflows.
  • Long-context – On GenAI4Science platform, this model runs with a minimum context window of 256k tokens — well suited for long documents, large codebases, and extended conversations.
  • Legacy – An older-generation model kept available for compatibility; newer models generally outperform it.
  • Frontier – A top-tier model competitive with the best currently available models. Temporarily available. Since we can run only one frontier model, it will be replaced promptly once a newer one arrives.

Select a model from the dropdown menu in the message box — pick a tag to narrow down the list.

DeepSeek

deepseek-r1:32b

Tags: Reasoning, Legacy

DeepSeek’s first-generation reasoning models (January 2025), achieving performance comparable to OpenAI-o1 across math, code, and reasoning tasks.

Distilled models

DeepSeek team has demonstrated that the reasoning patterns of larger models can be distilled into smaller models, resulting in better performance compared to the reasoning patterns discovered through RL on small models. The Qwen distilled models are derived from Qwen-2.5 series, which are originally licensed under Apache 2.0 License, and now finetuned with 800k samples curated with DeepSeek-R1.

The model weights are licensed under the MIT License. DeepSeek-R1 series support commercial use, allow for any modifications and derivative works, including, but not limited to, distillation for training other LLMs.

Google Gemma

Gemma is a family of open models built by Google DeepMind.

gemma4:31b

Tags: Vision, Code, Reasoning, Long-context

Gemma 4 models (April, 2026) are multimodal, handling text and image input and generating text output.

Gemma 4 introduces key capability and architectural advancements:

  • Reasoning – All models in the family are designed as highly capable reasoners, with configurable thinking modes.
  • Extended Multimodalities – Processes Text, Image with variable aspect ratio and resolution support (all models)
  • Diverse & Efficient Architectures – Offers Dense and Mixture-of-Experts (MoE) variants of different sizes for scalable deployment.
  • Optimized for On-Device – Smaller models are specifically designed for efficient local execution on laptops and mobile devices.
  • Increased Context Window – The small models feature a 128K context window, while the medium models support 256K.
  • Enhanced Coding & Agentic Capabilities – Achieves notable improvements in coding benchmarks alongside native function-calling support, powering highly capable autonomous agents.
  • Native System Prompt Support – Gemma 4 introduces native support for the system role, enabling more structured and controllable conversations.

The model list also offers gemma4-reasoning:31b. This is not a separate model: it is identical to gemma4:31b, but configured with reasoning (thinking mode) enabled by default. It is exposed as a convenience, so that users can easily access the reasoning functionality.

gemma3:27b

Tags: Vision, Legacy

Gemma 3 (March, 2025) is a lightweight, family of models from Google built on Gemini technology. The Gemma 3 models are multimodal - processing text and images - and feature a 128K context window with support for over 140 languages. They excel in tasks like question answering, summarization, and reasoning, while their compact design allows deployment on resource-limited devices.

The model has good knowledge of the Hungarian language.

Z.ai

glm-5.3

Tags: Code, Reasoning, Long-context, Frontier

Z.ai's GLM-5.3 ~743B MoE (39B active) frontier-class model (August 28, 2026 - 512k context). Temporarily installed for anyone to try. It is the most capable open-weights model for coding, with major gains on long-horizon agentic tasks.

It offers three reasoning effort levels driven by the reasoning_effort field; the default is max:

  • Think Max (default): omit reasoning_effort, or set "max" – deepest reasoning, for hard math, multi-step planning, and agentic tasks. Highest token cost.
  • Think High: "reasoning_effort": "high" – balanced depth and latency.
  • Think Low: "reasoning_effort": "low" – lightest reasoning, for simple Q&A, with the lowest latency and token cost.

More information: https://huggingface.co/zai-org/GLM-5.3

Meta Llama

Llama (Large Language Model Meta AI) is a family of large language models released by Meta AI starting in February 2023.

llama3.1:8b

Tags: Legacy

The upgraded version (July 23, 2024) of the 8B model is multilingual and has a significantly longer context length of 128K, state-of-the-art tool use, and overall stronger reasoning capabilities. This enables Meta’s models to support advanced use cases, such as long-form text summarization, multilingual conversational agents, and coding assistants.

llama3.3:70b

Tags: Legacy

The Meta Llama 3.3 (December 7, 2024) multilingual large language model is a pretrained and instruction tuned generative model in 70B (text in/text out). It offers similar performance compared to the Llama 3.1 405B model and is optimized for multilingual dialogue use cases.

It supports 7 languages in addition to English: French, German, Hindi, Italian, Portuguese, Spanish, and Thai.

llama4:scout

Tags: Vision, Legacy

Llama 4 Scout (April, 2025) is natively multimodal:

  • Input: multilingual text, image
  • Output: multilingual text, code

It is a 109B parameter MoE model with 17 billion active parameters and 16 experts, the best multimodal model in the world in its class and more powerful than all previous generation Llama models. Llama 4 Scout delivers better results than Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1 across a broad range of widely reported benchmarks.

Meta Muse

muse-glimmer:30b

Tags: Vision, Code, Reasoning, Long-context

Muse Glimmer 30b (August, 2026) is Meta's latest causal language model distilled from Muse Spark and purpose-built for autonomous agentic tasks on consumer hardware. The model integrates multi-step reasoning, reliable tool use, multimodal understanding, and failure recovery into a single model that runs locally without requiring cloud infrastructure or network access.

Mistral

Mistral AI is a French company specializing in artificial intelligence products.

mistral-small3.2:24b

Tags: Vision

Mistral-Small-3.2-24B-Instruct-2506 (June, 2025) is a minor update of Mistral-Small-3.1-24B-Instruct-2503.

Small-3.2 improves in the following categories:

  • Instruction following: Small-3.2 is better at following precise instructions
  • Repetition errors: Small-3.2 produces less infinite generations or repetitive answers
  • Function calling: Small-3.2’s function calling template is more robust

In all other categories Small-3.2 should match or slightly improve compared to Mistral-Small-3.1-24B-Instruct-2503.

OpenAI

OpenAI, Inc. is an American artificial intelligence organization founded in December 2015 and headquartered in San Francisco, California.

gpt-oss:120b

Tags: Reasoning

OpenAI’s gpt-oss:120b (August 5, 2025) achieves near-parity with OpenAI o4-mini on core reasoning benchmarks. It is designed for powerful reasoning, agentic tasks, and versatile developer use cases, and also performs strongly on tool use, few-shot function calling, and CoT reasoning.

The model is a Transformer which leverages mixture-of-experts (MoE) to reduce the number of active parameters needed to process input, activating 5.1B parameters per token.

It is trained on a mostly English, text-only dataset, with a focus on STEM, coding, and general knowledge.

https://openai.com/index/introducing-gpt-oss/

Qwen

Qwen is a family of large language models developed by Alibaba Cloud.

qwen3.8:27b

Tags: Reasoning, Code, Vision, Long-context

Qwen3.8 (August 14, 2026) is a 27b dense model delivering substantial gains across coding, professional work, research, and long-horizon agentic tasks.

  • Core Capabilities: Comprehensive improvements across coding, professional work, research, and long-horizon agentic tasks.
  • Agent Execution: Stronger autonomous planning and better handling of environment feedback, leading to more reliable end-to-end task completion.
  • Downstream Compatibility: Broader support for popular harnesses and development tools, making it easier to integrate into your existing stack.
  • Flexible Thinking Control: Thinking mode is on by default and can be disabled per request; reasoning depth can be tuned with reasoning_effort, and reasoning context from historical messages is retained via preserve_thinking.
  • Vision-Language Understanding: Native support for image and video understanding, from STEM diagrams and documents to hour-scale videos.

Vision models

These models process both text and image inputs:

Embedding models

The following embedding models are available:

Model Context Window
bge-m3:latest 8k tokens
embeddinggemma:latest 2k tokens
jina-embeddings-v5-text-small-retrieval:latest 16k tokens
nomic-embed-text:latest 8k tokens
nomic-embed-text-v2-moe:latest 512 tokens
qwen3-embedding:0.6b 16k tokens

These models are accessible only through the Performance endpoint. Contact us to request access.

Previously available models

The model lineup is regularly updated: older models are removed when newer ones take their place. The following models were previously available on the portal but can no longer be selected. If you require one of these models, please contact us.

Model Note
deepseek-r1:70b Larger distilled variant (DeepSeek-R1-Distill-Llama-70B)
gemma4:26bA4b Gemma 4 Mixture-of-Experts (MoE) variant
llama3.2:3b Small multilingual text-only model (September, 2024)
llama4:maverick Llama 4 Maverick, 400B parameter MoE model (April, 2025)
codellama:13b Code model built on Llama 2 (August, 2023)
mistral-large:123b Mistral Large 2 dense model (November, 2024)
devstral:24b Deprecated
devstral-small-2:24b Agentic coding model (December, 2025)
codestral:22b Mistral code model (May, 2024)
gpt-oss:20b Smaller OpenAI open-weight model (August, 2025)
qwen3.5:27b, 35bA3b, 122bA10b Qwen 3.5 multimodal family (February, 2026)
qwen3 Deprecated
qwen3-vl Deprecated
qwen3-coder:30b Agentic code model (July, 2025)