Lambda

AI infrastructure is the largest industrial buildout of our lifetime. Lambda is assembling the leadership team to match the opportunity ahead. Today, Lambda welcomes global infrastructure operator Michel Combes as CEO and former AT&T Communications CEO John Donovan as Chairman of the Board. Co-founder Stephen Balaban takes on the CTO role full-time, shaping the technology that will define the next decade of AI compute. Read the exclusive from @business: bloomberg.com/news/articles/…

@digitalix Deployed this literally while in an airplane on 4x B200’s

Heard some good things about Hy3, so ran some benchmarks and made a @LambdaAPI model card for it on 4xB200’s, enjoy! lambda.ai/inference-mode…

We broke down where GLM 5.2 lands and why the hard part isn't the model anymore. Read more: lambda.ai/blog/glm-5.2-a…

Intelligence is becoming a commodity. The bottleneck moves to infrastructure, to who can serve it, at what throughput, and at what cost. Running GLM 5.2 unquantized on-premises costs hundreds of thousands of dollars, and the throughput still won't match that of a real cluster.

Open-weight models hit a tipping point. GLM 5.2 runs as a real subagent in production. Not as a demo. Not as a benchmark screenshot. But as a near-frontier performance you can actually deploy.

Lambda has been recognized in the first Gartner® Magic Quadrant™ for Cloud AI Infrastructure, based on completeness of vision and ability to execute. It's also the first time Lambda has been recognized in one. This recognition reflects our focus: delivering against our commitments. Read the report: lambda.ai/gartner-magic-…

We're glad you asked 😉 Our pixel letters celebrate experimentation. Instead of polished and finished, they capture ideas still taking shape. They reflect the reality of AI and engineering work: build, test, break, and improve, again and again. Every breakthrough begins as an experiment.

@keval_sha @gpusteve cc @LambdaAPI in case you beat me to asking design ;)

does anyone know why lambda does this with the u e and o?

Frontier model prices keep climbing. Open models keep closing the gap. Lambda's @TheZachMueller is teaching a free session on when to switch and how to make it work. Part of @HamelHusain and @sh_reya's AI Product Engineering course on Maven. July 24. Free: maven.com/lls/21f487

@mattturck @stephenbalaban @LambdaAPI just got to hear it. wonderful must listen @stephenbalaban the original neocloud builder

Happy Fourth of July from the Lambda team.

Scaling first and debugging later is an expensive habit. @TheZachMueller does the reverse: he proves the job on the GPUs he already has, drives out every hidden wait, then promotes the exact same code to bigger GPUs. In this video, he trims a 500-million-parameter mixture-of-experts from 62 hours to 13 on an everyday four-GPU desktop. No exotic hardware. He shows where the idle seconds really come from, fixes only what moves the needle, and quits when the next tweak would cost more than it saves. Watch the talk and take the checklist you can rerun before you scale: youtube.com/watch?v=s_hSPB…

This is part of a broader vision of AI for Research: building AI researchers to automate research, one of the hottest areas in the field right now. This paper tackles one piece of it, the review process. Read more: arxiv.org/abs/2604.14261

Not replacing reviewers, but giving them concrete rubrics that turn vague assessments into specific, actionable critique.

Lambda research at ACL 2026. This week in San Diego, we're presenting ReviewGrounder, a system that makes peer review more substantive with LLMs. It outperforms GPT-4o by 135%.

Lambda keynote at ALVR, co-located with @aclmeeting (#ACL 2026) in San Diego. Our research team will cover published multimodal AI papers and walk through the infrastructure patterns behind them: how data moves, how training scales, and where the bottlenecks hide. July 3: lambda.ai/blog/lambdas-k…

As promised, more model cards for you: Kimi K2.7-Code and GLM-5.2! Both of which need zero introduction, as GLM-5.2 especially has taken the world by storm, but there's a new addition to our cards going in now: @NoumenaAI `code` usage directions! Kimi: lambda.ai/inference-mode… GLM: lambda.ai/inference-mode…

New model card up, @MiniMax_AI M3! (Working through the Colorado backlog) At 400B+ parameters, using the unquantized weights ends up needing a full HGX B200 (and I don't believe we can run the MXFP4 on hopper) Nice addition (on top of performance) is the multi-modality 😍

.@raais 2026, 10th edition. DeepMind on RL and robotics, Odyssey's live GoldenEye world model, ElevenLabs on serving at scale. 400+ researchers and builders in London. One of the few rooms where infrastructure and research collide. Thanks, @nathanbenaich.

Lambda's @chuanli11 and Amir Zadeh co-authored this one. 3D-DLP learns object-centric 3D representations from raw RGB-D and voxel inputs, with no supervision. Wins 9 of 10 RLBench tasks against matched baselines. Accepted at ICML 2026.

Autonomous by default. Human on escalation. the_lab.api is open source: lambda.ai/blog/what-happ…

.@claudeai Code + spare GPUs + an experiment tracker = a Gemma 4 model that learned Tetris in 2.5 days.