Somanshu Singla

AI Research @MicrosoftAI; ex-@essential_ai

[Mustafa Suleyman @mustafasuleyman]

We now have state-of-the-art pre-trained base models in the 3T-parameter weight class, built entirely in-house at MAI.

Post-training results coming soon -- stay tuned!

Lots of details in our technical blog post.

https://t.co/xcZ5uV4TAp

[Somanshu Singla @ssingla17]

@anlthms 🚀

[Somanshu Singla @ssingla17]

@_jainyash @timesofindia lets go!!

[Microsoft AI @MicrosoftAI]

MAI-Thinking-1: A powerful reasoning model developed from scratch that is competitive with models of similar size on STEM reasoning and coding tasks.

Our pre-training focused on a simple scaling emphasizing data-driven iterative improvements to our architecture and data. Our reinforcement learning (RL) framework is optimized for sustained log-linear climbs over many thousands of steps

We are openly sharing all technical details and learnings to build a transparent and science-driven approach to further development in AI

Read More: https://t.co/yK9Cd5loUd

[Mustafa Suleyman @mustafasuleyman]

Super excited to announce seven new world-class MAI models today. They represent what we consider a new era in AI designed to keep you in control and on the frontier.
First is our text foundation model, MAI-Thinking-1, exceptionally strong on reasoning and SWE tasks.
- It’s a 35B active parameter MoE with a 256K context window. Independent human raters on Surge prefer it for overall quality in blind side-by-sides versus Sonnet 4.6, and it’s achieved 97% on AIME 2025, the key measure of its general-purpose reasoning abilities.
- It's at 53% on SWE Bench Pro, placing it right alongside Opus 4.6 on one of the toughest coding benchmarks.
- And since we co-designed our models with our own silicon, MAI-Thinking-1 is optimized on our MAIA 200 chip. Benchmarking head-to-head against the GB200, we see 30% better performance per dollar as well as a 1.4x performance-per-watt gain when running our MAI models on the MAIA 200 end-to-end.

Next is MAI-Image-2.5 and its Flash variant. Two super strong models now at #2 on the leaderboards, surpassing the score of Nano Banana 2 on image editing.

Last for now is MAI-Code-1-Flash, our new inference efficient coding model, especially tuned for VS Code and GitHub Copilot CLI.
- Code-1-Flash achieves 51% on SWE Bench Pro, despite having just 5B parameters, putting it closer to Haiku in size but cheaper in cost.

All of this is the foundation for Microsoft Frontier Tuning. It lets you customize our models to create custom, company-specific agents that only you control. You can make our model, your model. Your data. Your agents. Your moat.

Early adopters are already seeing a difference. When we tuned our models for McKinsey’s tasks, MAI delivered the highest win rate, outperforming GPT-5.5 on quality, while being 10x lower on cost.

Also really excited to be collaborating with the amazing team at Mayo Clinic to jointly train a new frontier AI model for healthcare.

Our announcements today mark another milestone on the road to humanist superintelligence. You can learn more and about our other new models in our latest blog: https://t.co/v65eop5Ixq

[Somanshu Singla @ssingla17]

Amazing work by @gauri__gupta @RitvikKapila !

Self-improving AI systems are the future

[Gauri Gupta @gauri__gupta]

We @neosigmaai @RitvikKapila are building the future of self-improving AI systems! By closing the feedback loop between production data and system improvements, we help teams capture failures, convert them into structured evaluation signals, and use them to drive continuous improvements in agent behavior.

We show how our system works on Tau3 bench across retail, telecom, and airline domains. Agent performance on the validation set (with a fixed underlying model, GPT5.4) improves from 0.56 → 0.78 (~40% jump in accuracy).

[Somanshu Singla @ssingla17]

Very exciting result!
We achieve this without any telecom specific training 🚀

[Essential AI @essential_ai]

Rnj-1’s performance is especially good in correctness and abstention in its weight class, which are the two most important metrics for this work.

[Vipul Ved Prakash @vipulved]

The OSS coding model that @ashVaswani and team at @essential_ai have been working on over the last year is now out and available to use on @togethercompute APIs and playground.

It's particularly exciting that @essential_ai is doing fundamental work in pre-training and post training and publishing open weights!

https://t.co/mTKOV49jOz
https://t.co/di9a968ZLN

[Yupp @yupp_ai]

📢 New Model Drop: Rnj-1 Instruct is now live on Yupp!

The first flagship model from @essential_ai, this open-source model is built to excel at math problem-solving and scientific reasoning.

We checked it out with some prompts:

[Ashish Vaswani @ashVaswani]

We made new research bets on modeling program behavior that accelerated the model’s coding abilities. Tremendous advances made by @_jainyash @DevaanshGupta1 @cadarsh_335 @ssingla17 @anlthms @_saurabh.

[Essential AI @essential_ai]

Today, we’re excited to introduce Rnj-1, @essential_ai's first open model; a world-class 8B base + instruct pair, built with scientific rigor, intentional design, and a belief that the advancement and equitable distribution of AI depend on building in the open.

We bring American open-source at par with the best in the world.

[Somanshu Singla @ssingla17]

Super proud to finally share Rnj-1 with the world!

It took a ton of work (and a lot of fun) to build, and I’m incredibly excited for everyone to try it out.

I’m confident Rnj-1 won’t disappoint :)

[Ashish Vaswani @ashVaswani]

We are beyond thrilled to share our first flagship models, Rnj-1 base and instruct 8B parameter models. Rnj-1 is the culmination of 10 months of hard work by a phenomenal team, dedicated to advancing American SOTA OSS AI.

Lots of wins with Rnj-1.

1. SWE bench performance close to GPT 4o.
2. Tool use outperforming all comparable open source models.
3. Mathematical reasoning (AIME’25) nearly at par with GPT OSS MoE 20B.

….

[Essential AI @essential_ai]

[1/2] We at Essential are driven by mission to advance fundamental research guided by first principles, rigor and sharing research openly.

[Harman Singh @ ICML 🇰🇷🇰🇷 @Harman26Singh]

🚨 New @GoogleDeepMind paper

𝐑𝐨𝐛𝐮𝐬𝐭 𝐑𝐞𝐰𝐚𝐫𝐝 𝐌𝐨𝐝𝐞𝐥𝐢𝐧𝐠 𝐯𝐢𝐚 𝐂𝐚𝐮𝐬𝐚𝐥 𝐑𝐮𝐛𝐫𝐢𝐜𝐬 📑
👉 https://t.co/oCk5jGNYlj

We tackle reward hacking—when RMs latch onto spurious cues (e.g. length, style) instead of true quality.
#RLAIF #CausalInference

🧵⬇️

[Somanshu Singla @ssingla17]

We didn't see clear wins on grokking with Muon.
Curious to hear if others have tried muon on grokking and their findings!

[Essential AI @essential_ai]

[1/5]

We have a quick update to share, which contradicts our hypothesis regarding the abilities of Muon and Adam vis-a-vis Grokking.

[Ritvik Kapila @ritvikkapila]

#1 trending on @huggingface letsgoooo!

@essential_ai 🥇

[Ashish Vaswani @ashVaswani]

Check out our latest research on data. We're releasing 24T tokens of richly labelled web data. We found it very useful for our internal data curation efforts.

Excited to see what you build using Essential-Web v1.0!

[Somanshu Singla @ssingla17]

Check out our work on "Practical Efficiency of Muon for Pretraining" !

[Ashish Vaswani @ashVaswani]

Please check out our thorough study on the advantages of Muon. Second-order optimization is a promising path to more efficient LLM pretraining.

[Darsh J Shah @darshj_shah]

Reflection improves with pre-training. Triggers like ‘Wait,’ enhance this dramatically!

[Somanshu Singla @ssingla17]

Excited to share our work at Essential!
While most recent research has focused on how reflection develops during reinforcement learning.
We show that it actually begins to emerge much earlier — during model’s pre-training!

[Ashish Vaswani @ashVaswani]

Reinforcement learning has shown success in eliciting reflection from LLMs, but what if this capability actually manifests earlier in pre-training? We investigated this question and our results are surprising 👇

[1/4]