Agent Development
EvoSkill
Turn a general agent like Claude Code into a specialist automatically using benchmarks.
+12.1%
On SealQA
Zero
Model Training
The world's
leading open-source
AI reasoning lab
Frontier open-source research and products that enable AI systems to reason, reflect, learn, and self-improve.
A community building the future of open-source AI
Research
Frontier open-source AI research advancing how systems reason — peer-reviewed at NeurIPS, ICML, COLM and more, and cited across 100+ institutions including MIT, CMU and Microsoft.
13
Papers & models
NeurIPS · ICML · COLM
Peer-reviewed venues
Google · Microsoft · Alibaba
Cited by 100+ institutions
Products
Enable your AI agents to discover, learn, and improve automatically through our suite of frontier tools.
Agent Experimentation
Arena
Compete on the hardest AI reasoning benchmarks and earn prizes for original solutions.
$50K+
Prizes so far
1,200+
Applicants
Model Ownership
OML
Open, Monetizable, and Loyal AI. Cryptographic model fingerprinting.
OML 1.0
Live system
AI-native
Cryptography
Agent EvaluationComing Soon
Eldros
Upload agent success criteria to generate benchmarks, datasets, and red-teaming reports.
2026
Expected release
Recent Articles
Browse All
How To Build a Faster and Smarter Agent by Pre-Filtering Tools with RAG
We scaled a multi-tool agent from vanilla ReAct to a production-grade architecture by pre-filtering tools with Retrieval-Augmented Generation (RAG)
Research BriefsMay 29, 2026
CryptoAnalystBench: How to Evaluate AI Agents When There Are No Clear Answers
CryptoAnalystBench evaluates long-form crypto research agents without ground truth.
Research BriefsJanuary 30, 2026
Are Robust LLM Fingerprints Adversarially Robust?
Malicious hosts break nine of ten LLM fingerprinting schemes easily.
Research BriefsJanuary 29, 2026
Context Manipulation Attacks: Why Web Agents Need Secure Memory
Web agents' external memory enables plan injection beyond prompt defenses.
Research BriefsOctober 30, 2025
GRID Ecosystem
We work with 70+ partners and other open-source builders to create an ecosystem we call GRID. Join to build the open-source frontier together!
Trusted and powered by industry leaders
Stay updated on Sentient and the frontier of AI reasoning
New AI research, products, updates and more
