←
RahimEntezari900 posts
RahimEntezari avatar

RahimEntezari

@rahiment

Senior Research Scientist @wayve_ai, ex-StabilityAI

◷ Joined August 2017▧ 39 Photos and videos
Robin Rombach avatar
⟳ RahimEntezari retweeted
Robin Rombach@robrombach·Nov 12

FLUX-mimic goes brrr 🔥

Today we’re announcing a joint research project with @BMW Group on video-action models for industrial manipulation. We trained FLUX-mimic on automotive tasks using production parts and fixtures: assembly, insertion, kitting and cable routing. On our dexterous robot platform, it achieved success rates of up to 100% and consistently outperformed other models, even with 4x less training data. Collecting task-specific data is the biggest cost of bringing learned policies into factories today. From day one, we’ve believed that working closely with industrial partners on real production data is the only path to deployment.

1:03
Behnam Neyshabur avatar
⟳ RahimEntezari retweeted
Behnam Neyshabur@bneyshabur·Nov 12

We’ve raised a $1B Series A, led by @kleinerperkins and @clearlake. Capital is a means, not an end. Our goal is radical leverage for scientists: One person doing what took a team. A team doing what took an institution.

Today we're announcing our $1B Series A, led by @kleinerperkins and @clearlake. We believe making frontier AI R&D more accessible is the most direct path to AI that solves humanity's most pressing problems across science and technology.

Post attachment
Mirendil avatar
⟳ RahimEntezari retweeted
Mirendil@mirendil·Oct 26

Introducing Fern, Mirendil’s first model: pretrained from scratch, with up to 10 million tokens of context. That’s ~0.5M lines of code or days of an agent’s work - all in context at once. Built for high-volume agentic work in AI R&D: auditing training data, debugging training runs from logs, and reading agent traces end to end. Fern’s next-token prediction keeps improving as context grows to 10M tokens. Fern 0.1 is Mirendil-Base after a short round of post-training. It’s the first and smallest of three models, with two larger ones now being pretrained. We’ll ship frequent updates and make the model available soon.

Post attachment
Ramin avatar
⟳ RahimEntezari retweeted
Ramin@ramin_m_h·Oct 25

what a day at the track in Austin! congratulations to Kimi Antonelli for a well deserved win 🏎️ Incredibly grateful of @MercedesBenz for the invite!

Post attachmentPost attachmentPost attachment
0:16
Behnam Neyshabur avatar
⟳ RahimEntezari retweeted
Behnam Neyshabur@bneyshabur·Oct 26

Introducing Fern, Mirendil’s first model: pretrained from scratch, with up to 10M tokens of context! That’s ~0.5M lines of code (roughly Megatron-LM scale) or days of an agent’s work, all in context at once. Its first version, Fern 0.1, had a short round of post-training 🚢

Post attachment
Mirendil avatar
⟳ RahimEntezari retweeted
Mirendil@mirendil·Oct 21

Today, we're sharing a major milestone in our mission to democratize frontier AI development: Pretraining at scale without a pretraining team. We set out to answer an ambitious question: Can we significantly reduce the time, human expertise, and compute needed to pretrain a competitive foundation model from scratch? The result is Mirendil-Base, our first pretrained coding model, built by one researcher in just 3.5 months with less than $10M in compute, including all research, experimentation, and the final training run. Our approach is already showing promising results as we scale to even larger models. This is an important step toward a future where a single researcher, empowered by AI, can accomplish what once required an entire research organization. As these capabilities improve, we believe frontier AI development can become accessible to far more people and teams. More to come soon, including independent evaluations, post-training benchmarks, and API access.

Post attachment
Behnam Neyshabur avatar
⟳ RahimEntezari retweeted
Behnam Neyshabur@bneyshabur·Oct 23

We asked Baseten to independently evaluate @mirendil’s base models. We shared our model endpoints so they could evaluate them on their own private codebases. Mirendil-Base-Large ended up being the best of all the models they tested! Thank you, @baseten, for being a great partner!

We ran @mirendil's new base model on Baseten's private codebases; that is, code it's never seen. Mirendil-Base-Large had the lowest bits per byte (a tokenizer-independent measure of perplexity) of any model we tested. We're excited to be the inference partner to serve Mirendil models. Congrats to the Mirendil team on this launch!

Post attachment
Behnam Neyshabur avatar
⟳ RahimEntezari retweeted
Behnam Neyshabur@bneyshabur·Oct 21

We formally announced Mirendil less than 4 months ago. Today, I'm excited to share an important milestone for: Pretraining at scale without a pretraining team. Mirendil's mission is to solve Recursive Self-Improvement (RSI) and use it to democratize frontier AI development, ultimately accelerating science and technology. At Mirendil, when we think about RSI, we think holistically about the company as an intelligent entity that self-improves toward its goal. From that perspective, the most powerful notion of acceleration is reducing the time and resources (number of human experts, compute, and other costs) needed to achieve ambitious goals. Given that pretraining is perhaps the most important yet most challenging part of frontier AI development, we decided to run an ambitious controlled experiment: Starting fully from scratch, can we significantly reduce the time, number of human experts, and compute resources needed to go all the way to successfully pretraining a strong coding model at scale? Mirendil-Base is the result of this experiment. Our first pretrained model was built by one researcher in 3.5 months, with less than $10M in compute. This includes all the research and small-scale experiments it took to get there, as well as the final training run. The recipe holds up as the models get bigger. For reference, Mirendil-Base is in the 100B-500B parameter range, Mirendil-Base-Large in the 500B-2T range, and Mirendil-Base-XL in the 2T-10T range. We're beginning to see a world where a single researcher, empowered by AI, can take on work that once required an entire research organization. As these systems improve, the bottleneck shifts from the size of a team to the ambition and judgment of the people directing them. This has profound implications for who gets to build AI and, ultimately, who gets to push the frontiers of science and technology forward. We’ll share independent evaluations, post-training benchmarks and API-access sooner than you think so stay tuned!

Post attachment
Ramin avatar
⟳ RahimEntezari retweeted
Ramin@ramin_m_h·Oct 11

folks, just go with the d1 models for your decision use-cases! its just getting started 🌊 Incredible explainer @TheTuringPost 🙏🏻✨ Openrouter: openrouter.ai/liquid/d1 HF: huggingface.co/LiquidAI/d1-3… HF: huggingface.co/LiquidAI/d1-o…

stumbled over this fantastic video by @TheTuringPost on our open d1 models. this week we released our open d1 decision models and this video discusses: • why fast & multimodal decision models are important • why building on LFM2.5 models enables fast decisions • what use cases fast & multimodal decision models unlock video: youtube.com/watch?v=EjX3OQQN… ps: so glad to see i'm not the only one struggling with that hand flappy demo. it's so hard.

Ramin avatar
⟳ RahimEntezari retweeted
Ramin@ramin_m_h·Aug 6

LFM vs DeepSeek Flash on agentic tool calls

New LFM2.5-2.6B hits DeepSeek-V4 level on tool calling and runs 3.7x faster! We ran @liquidai 's new LFM2.5-2.6B against DeepSeek-V4-Flash on one box with 4x RTX 5090. Both got the same three jobs, and each one only completes if the model fires every tool call Topics: -weather and local time in six cities -one budget into six currencies -four hotels checked and booked for one date Outputs: LFM2.5-2.6B: 35/35 tool-calls, 366 tok/s, 19s DeepSeek-V4-Flash: 35/35 tool-calls, 77 tok/s, 70s Both models made all 35 calls, and both fired twelve of them in one turn. LFM was trained for agents, and tool calling is where that shows. Strong result for a model you can run right on your phone

0:25
TechCrunch avatar
⟳ RahimEntezari retweeted
TechCrunch@TechCrunch·Aug 6

Mirendil has signed a $100 million-plus Google Cloud partnership to expand its compute infrastructure, powering research into self-improving AI systems designed to accelerate scientific discovery and AI development. spr.ly/6013B18mb7

Behnam Neyshabur avatar
⟳ RahimEntezari retweeted
Behnam Neyshabur@bneyshabur·Aug 6

@Google has been a reliable partner for @mirendil since day one. Today, we are excited to announce a multi-year partnership to power the next generation of self-accelerating AI!

Today, we're announcing a multi-year partnership with @Google to power the next generation of self-accelerating AI! Mirendil will scale its training, inference, and AI research workloads on @googlecloud 's AI Hypercomputer — drawing on both TPUs and full-stack NVIDIA AI Infra.

Post attachment
Robin Rombach avatar
⟳ RahimEntezari retweeted
Robin Rombach@robrombach·Jul 21

0:21