Live · 2026
MAGIBU RESEARCH GROUP / İSTANBUL · TR

Building foundation models for
AI sovereignty.

Data sovereignty is national sovereignty.

Magibu turns Turkish foundation and embedding models into secure, measurable AI systems that run inside your organization. We invite every organization that values its data to join this journey.

Magibu Q3 · Foundation ModelTR-MMLU · 6200+ questionsembeddingmagibu-200mTurkish Morphological TokenizerTR-MTEB · 26 datasetsOpen sourceMagibu Q3 · Foundation ModelTR-MMLU · 6200+ questionsembeddingmagibu-200mTurkish Morphological TokenizerTR-MTEB · 26 datasetsOpen sourceMagibu Q3 · Foundation ModelTR-MMLU · 6200+ questionsembeddingmagibu-200mTurkish Morphological TokenizerTR-MTEB · 26 datasetsOpen source
Token efficiency

Fewer tokens mean lower cost and lower energy.

When Turkish text is represented closer to the structure of the language, the same meaning can travel through fewer pieces. That chain connects directly to inference cost, energy use, and secure systems that can run inside the organization.

01

Fewer tokens

A tokenizer approach that treats Turkish suffixes and meaning units more carefully.

02

Lower cost

Less processing for the same text and more predictable API/infrastructure cost.

03

Lower energy

Shorter context and less compute create a measurable energy advantage.

04

Data stays inside

Efficient models make on-prem and private cloud deployments more practical.

Try your own sentence

Turkish tokenizer comparison

Open demo ↗

If the demo does not load, open the comparison on Hugging Face.

§ 01Live Demos

Try what we have shipped.

Test our models, tokenizers, and embeddings in live demos. Every link points to an open-source release or a published product.

MODEL

Magibu Q3

Our foundation model for Turkish text generation and understanding. Try it in the live chat interface.

Start chatting
BENCHMARK

TR-MMLU

Open MMLU list with 6200+ questions across 57 domains. Compare Turkish models side by side.

Open full list
TOKENIZER

Turkish Tiktokenizer

Morphological tokenizer built for Turkish. Compare suffix handling and token efficiency live.

Open demo
DEMO

embeddingmagibu-200m

Turkish-first embedding model. Live semantic similarity demo.

Open demo
BENCHMARK

TR-MTEB Scoreboard

Open embedding leaderboard across 26 datasets and 6 tasks. Compare models side by side.

Scoreboard
§ 02Products

Retrieval Platform three layers.

A full stack from embedding to AI system to in-house deployment. Take one layer, or run all three together.

I Live Demo

Magibu
Embed API

An OpenAI-compatible, high-performance embedding API optimized for Turkish and underrepresented languages. Superior semantic representation with long-context support.

api.magibu.ai/embeddings
API Platform ↗
II Pilot

Magibu
Search Kit

Production-ready retrieval infrastructure for AI applications. Automated document ingestion, semantic chunking, vector database connectors, and query evaluation tools.

+ Qdrant · pgvector · Weaviate
III Enterprise

Magibu
Private AI

An isolated AI architecture running fully on-premises or within your private cloud. Domain-adapted search, source-grounded answers, SSO/LDAP integration, and security audit logs.

Docker / Kubernetes / On-Prem
IV Live Demo

Magibu
Q3 Foundation

Our foundation model optimized for Turkish text generation and comprehension. Integrated with enterprise security standards during pilot deployments.

§ 03Entry Product

Measure first, then deploy.

Magibu Retrieval Audit is a 2-week measurement package before pilot. We compare models and architectures on your data together.

Magibu Retrieval Audit · 2 weeks
"Let's measure which model works better on your documents; then deploy a secure in-house AI system."

Not a sales pitch - a way of working. You can't be sure of progress in a system you haven't measured; we set direction with metrics like recall, precision, and MRR, not guesswork. Before buying a large system, we measure on your own data. Most organizations skip this step and come back months later. We start here.

Audit2 weeksFixed fee · 1 department
AI Pilot4 weeks1–3 data sources · working demo
Private Deployment8–12 weeksOn-prem · SLA · SSO/audit
01

Data Sampling and Analysis

We select a representative subset of your documents together. Data stays inside your environment.

02

User Test Scenarios

30–100 real user questions with expert-labeled correct passages.

03

Model & Architecture Benchmark

Magibu, OpenAI, Cohere, Voyage, and E5 measured on the same data. Chunking strategies compared side by side.

04

Comprehensive Metrics Report

recall@5, precision@10, MRR, nDCG@10, and latency. Top 5 wins + 5 critical failures with case studies.

05

Topology Recommendation

One-page technical and financial rationale for model, chunking, reranker, vector DB, and deployment topology.

06

Roadmap & Decision

Continue or stop for pilot. Audit delivers value on its own; not required before pilot.

Scientific assurance

Research output turned into product evidence.

Our technology is built on open benchmarks, doctoral research, and peer-reviewed or openly published work. Instead of naming committee members on the homepage, we surface verifiable publications and model outputs.

AI Safety2026

Benchmarking Prompt Injection Attacks on LLMs

Applied Sciences · accepted

Open paper ↗
Benchmark2024

Setting Standards in Turkish NLP: TR-MMLU for Large Language Model Evaluation

arXiv:2501.00593

Open paper ↗
Tokenization2025

Tokenization Standards for Linguistic Integrity: Turkish as a Benchmark

arXiv:2502.07057

Open paper ↗
Embedding2026

Adapting Multilingual Embedding Models to Turkish via Cross-Lingual Tokenizer Surgery and Offline Distillation

arXiv:2605.29992

Open paper ↗
Medical LLM2025

Healthcare-Focused Turkish Medical LLM

ACM TALLIP

Open paper ↗
§ 04Vision

Three arms, one evidence culture.

Company ships systems inside organizations; Academy teaches the work; Community grows open measurement and open science.

CompanyCommercial

Products
behind your firewall.

We build AI systems for security-sensitive organizations, starting with measurement and moving toward private cloud or on-prem deployment.

  • 01
    Retrieval AuditModel and architecture comparison on your own data
  • 02
    Retrieval PlatformEmbed API · Search Kit · Private AI
  • 03
    Private DeploymentOn-prem / private cloud · SSO · audit logs
  • 04
    Enterprise TrainingArchitecture design, data security, and model evaluation
AcademyTraining

Knowledge
turned into practice.

We turn LLM, embedding, RAG, and evaluation expertise into production-oriented training where participants build instead of only watching.

  • 01
    Builder-first programsLearning through GitHub and Hugging Face outputs
  • 02
    Enterprise trainingAI architecture for technical teams and executives
  • 03
    Certificate disciplineVerifiable achievement and portfolio, not attendance only
  • 04
    Open-source connectionTraining outputs connect back to community projects
CommunityOpen Source

Open measurement
and open science.

Our community branch grows open model, tokenizer, dataset, and benchmark work for Turkish and low-resource languages.

  • 01
    Transparent DevelopmentGitHub Issues + PRs · open contribution flow
  • 02
    Open BenchmarkTR-MTEB · TR-MMLU · domain-specific eval kits
  • 03
    Models & DataOpen model and dataset work on Hugging Face
  • 04
    Community EventsMeetups, hackathons, webinars, and the weekly digest
§ ContactEnterprise

applications and partnerships.

Apply for a Retrieval Audit, API access, investment, or research collaboration. Our team will respond within the shortest possible time.

"Data sovereignty is national sovereignty."
  • dev.magibu.ai · Embedding API
  • TR-MTEB · 26 datasets
  • On-prem / Private AI

Enterprise pilot, API access, investment, and partnership requests require a verified Google account.

Checking session…
Interest area *
✦ ✦ ✦
MMagibu is a research-driven technology group building AI infrastructure for Turkish and underrepresented languages.
Global models treat Turkish as secondary. We treat it as a primary research and product language.
Measurement sits at the foundation of everything: without measuring, you can't be sure progress is heading the right way. The community produces open measurement and science; the company turns that output into secure, scalable products for organizations.
Three arms, one vision: produce evidence, teach the work, and ship evidence as product.
Data sovereignty is national sovereignty. Under this sentence sit published benchmarks, open models, and live demos.