AI21 builds market-leading models for the enterprise. Our Foundation Model, Jamba, is a breakthrough SSM-Transformer model that is the first production-grade Mamba-based model delivering best-in-class quality and performance. AI2's Task-Specific Models are smaller, specialized models purpose-built to solve key enterprise challenges. Our Contextual Answers and Summarization TSMs offer unprecedented efficiency and cost-savings on core enterprise use cases.
339 signals in the last 90 days.
The order this company entered each market — oldest to newest
Sign up to see the full signal feed
How AI21 Labs's international expansion compares to its competitors
AI21 LabsYouBars show relative size across the companies compared.
Columns5| Company | Markets | New markets | Momentum | Activity | Expansion score |
|---|---|---|---|---|---|
 AI21 LabsYou | 1 | 0 | -97% | 339 | 31 |
DevOps Engineer
View postingTel Aviv-Yafo, Tel Aviv District, Israel · last mo.
Workplace Operations Team Lead (Maternity leave replacement)
View postingTel Aviv-Yafo, Tel Aviv District, Israel · 3mo ago
Executive Assistant- Maternity Leave Replacement
View postingTel Aviv-Yafo, Tel Aviv District, Israel · 3mo ago
I'm proud to share a new state-of-the-art result on the SWE-Rebench coding benchmark: a 60.9% issue resolve rate on issues from December 2025 to March 2026, beating the previous best.If you've used a coding agent like Claude Code or Cursor, you've probably noticed the pattern: it reads the issue, searches the repo to find the relevant files, then tries to solve. Research first, then solve. We tried that approach and went further - added a dedicated research step using a focused agent to gather ...
Ended
Caching is every eval engineer's best friend. Until your agent uses best-of-N sampling.But caching was built for a deterministic world: same input, same output, every time. LLMs, which are non-deterministic by nature, don’t play by these rules. Especially when we’re dealing with an agent that uses best-of-N sampling to improve accuracy.In the latest Labs in Front piece, AI21 Labs software engineer Guy Lichtenfeld tackles the question: How can we reproduce variance? In other words, how do we pres ...
EndedWhy your agent evals may be completely wrong
EndedAI21 Labs introduces Maestro, an AI planning system that improves LLM output quality by up to 50%, simplifying error mitigation and task management.
AI21 introduces Maestro, an AI planning system that improves accuracy of AI models by up to 50% on complex tasks.
Google introduces new tools for Vertex AI to build and deploy multi-agent AI systems, including Agent Developer Kit, Agent Garden, and Agent Engine.
Unlock verified emails & direct dials
Free · No credit card needed
Quick answers about AI21 Labs's business, markets, and growth signals.
Pubrio glocalizes business, people, and intent data from 50+ local sources into one global graph, giving AI and revenue teams full‑market visibility.
© 2026 Pubrio. All rights reserved.
