Skip to main content
Rase AI Intelligence Newsletter

Explore AI developments

Search published AI tools, models, robots, and research. Filter by organization, geography, milestones, and evidence.

Filters and sorting
Filter by milestone dates
Announcement
First access
Commercial release
Clear filters

69 profiles

Models & assistants

Viking

Silo AI (AMD) and University of Turku

It generates text in the language settings described by LumiOpen.

Medium confidenceEvidence checked

Models & assistants

EuroLLM

UTTER consortium (Unbabel, IST Lisbon, Edinburgh et al.)

It generates text for languages described in its model documentation.

Medium confidenceEvidence checked

Models & assistants

Salamandra

Barcelona Supercomputing Center

A multilingual family of language models from the Barcelona Supercomputing Center, with base and instruction-tuned checkpoints in several sizes.

Medium confidenceEvidence checked

Models & assistants

Teuken-7B

OpenGPT-X (Fraunhofer et al.)

A multilingual language model from the OpenGPT-X project, trained for text tasks across the European Union’s official languages.

Medium confidenceEvidence checked

Models & assistants

Bielik v2

SpeakLeash and ACK Cyfronet AGH

An instruction-tuned generative text model developed by SpeakLeash and the Academic Computer Centre Cyfronet AGH, with a focus on Polish-language use.

Medium confidenceEvidence checked

Models & assistants

PLLuM

Polish Ministry of Digital Affairs consortium

A family of language models optimized for Polish as well as other Slavic and Baltic languages, with additional English training data.

Medium confidenceEvidence checked

Models & assistants

Lucie-7B

OpenLLM-France (LINAGORA et al.)

A multilingual language model from the OpenLLM-France initiative, with its technical report documenting the model and associated training dataset.

Medium confidenceEvidence checked

Models & assistants

InkubaLM

Lelapa AI

A compact language model from Lelapa AI researchers for languages with limited training resources, evaluated on tasks such as translation and question answering.

Medium confidenceEvidence checked

Models & assistants

Sabiá

Maritaca AI

A family of language models developed for Portuguese by Maritaca AI, with the listed checkpoint accepting text input and returning text.

Medium confidenceEvidence checked

Models & assistants

Param-1 (BharatGen)

BharatGen (IIT Bombay consortium)

A bilingual language model from BharatGen trained in Hindi and English, with an instruction-tuned checkpoint described in its model card.

Medium confidenceEvidence checked

Models & assistants

Krutrim-2

Krutrim (Ola)

An instruction-tuned language model from the Krutrim team, built on the Mistral NeMo architecture and trained on material including Indian languages and code.

Medium confidenceEvidence checked

Models & assistants

Kimi K2

Moonshot AI

A mixture-of-experts language model from Moonshot AI designed for coding, tool use and other agent tasks.

Medium confidenceEvidence checked

Models & assistants

GLM-4.5

Z.ai (Zhipu AI)

A mixture-of-experts foundation model from Z.ai designed for agent workflows that involve reasoning, coding and tool use.

Medium confidenceEvidence checked

Models & assistants

MiniMax-M1

MiniMax

A MiniMax reasoning model that uses a mixture-of-experts architecture and hybrid attention designed to handle long inputs.

Medium confidenceEvidence checked

Models & assistants

Hunyuan-Large

Tencent

A Tencent language model using a mixture-of-experts architecture, with the technical report describing its design for text generation.

Medium confidenceEvidence checked

Models & assistants

InternLM3

Shanghai AI Laboratory

An InternLM text model published in an instruction-tuned checkpoint, designed for conversation, reasoning and coding tasks.

Medium confidenceEvidence checked

Models & assistants

Seed-OSS-36B

ByteDance Seed

A family of ByteDance Seed language models suited to long inputs, with a developer-described way to control the amount of reasoning.

Medium confidenceEvidence checked

Models & assistants

Step-3

StepFun

A StepFun multimodal reasoning model that combines image and text inputs and uses a mixture-of-experts architecture.

Medium confidenceEvidence checked

Models & assistants

ERNIE 4.5 (open weights)

Baidu

A Baidu model family using mixture-of-experts designs, with developer documentation describing text and multimodal variants.

Medium confidenceEvidence checked

Models & assistants

MiMo-7B

Xiaomi

A Xiaomi language model designed for reasoning tasks including mathematics and coding, with its development described in a published technical report.

Medium confidenceEvidence checked

Models & assistants

DeepSeek-R1

DeepSeek

It generates text responses designed to handle multi-step reasoning tasks.

Medium confidenceEvidence checked