Menlo Research
@menloresearch.bsky.social
Menlo is an open R&D lab building the brain for robots.
Website: https://menlo.ai/
Community: https://discord.gg/Exe46xPMbK
Website: https://menlo.ai/
Community: https://discord.gg/Exe46xPMbK
The hips use a F.A.R. layout: Flexion, Abduction, Rotation. All actuators line up on one plane. It's a clean way to explore stride, balance, and motion efficiency.
October 23, 2025 at 4:17 AM
The hips use a F.A.R. layout: Flexion, Abduction, Rotation. All actuators line up on one plane. It's a clean way to explore stride, balance, and motion efficiency.
We keep the core frame flexible so we can change the form or control setup easily. Nothing here is final, it's a testbed that evolves as we do.
October 23, 2025 at 4:17 AM
We keep the core frame flexible so we can change the form or control setup easily. Nothing here is final, it's a testbed that evolves as we do.
It's less about perfect structure and more about learning what works. We're tweaking joints, testing control, breaking parts, and fixing them the next day. Every version walks a bit differently.
October 23, 2025 at 4:17 AM
It's less about perfect structure and more about learning what works. We're tweaking joints, testing control, breaking parts, and fixing them the next day. Every version walks a bit differently.
We're going to have a pair of humanoid robot legs running around without a body soon.
We started building them because the humanoids you can buy aren't made for experiments, so we're building something we can actually experiment with.
We started building them because the humanoids you can buy aren't made for experiments, so we're building something we can actually experiment with.
October 23, 2025 at 4:17 AM
We're going to have a pair of humanoid robot legs running around without a body soon.
We started building them because the humanoids you can buy aren't made for experiments, so we're building something we can actually experiment with.
We started building them because the humanoids you can buy aren't made for experiments, so we're building something we can actually experiment with.
Rick's got a cool new head.
October 3, 2025 at 11:12 AM
Rick's got a cool new head.
Rick is hailing a cab.
October 3, 2025 at 2:49 AM
Rick is hailing a cab.
Rick is walking around tech stores.
October 2, 2025 at 8:28 AM
Rick is walking around tech stores.
Have your own version of Perplexity Pro's Deep Research with Jan v1, the open-source alternative.
Cookbook: jan.ai/post/jan-v1...
Quick reminder to make Jan v1 ready for search:
- Turn on Experimental Features
- Add a search API in MCP Server settings
- Enable tool use for the model
Cookbook: jan.ai/post/jan-v1...
Quick reminder to make Jan v1 ready for search:
- Turn on Experimental Features
- Add a search API in MCP Server settings
- Enable tool use for the model
August 22, 2025 at 12:26 PM
Have your own version of Perplexity Pro's Deep Research with Jan v1, the open-source alternative.
Cookbook: jan.ai/post/jan-v1...
Quick reminder to make Jan v1 ready for search:
- Turn on Experimental Features
- Add a search API in MCP Server settings
- Enable tool use for the model
Cookbook: jan.ai/post/jan-v1...
Quick reminder to make Jan v1 ready for search:
- Turn on Experimental Features
- Add a search API in MCP Server settings
- Enable tool use for the model
We're in Rotterdam for Interspeech 2025. If you're attending, feel free to stop by and say hello!
August 21, 2025 at 7:09 AM
We're in Rotterdam for Interspeech 2025. If you're attending, feel free to stop by and say hello!
"Hey folks - I left my head at home. If you find it, please tell it I'm not mad, just disappointed."
August 8, 2025 at 9:16 AM
"Hey folks - I left my head at home. If you find it, please tell it I'm not mad, just disappointed."
Meet Rick, the Unitree G1 we're hacking in the lab. When he's not sneaking out for K-TV, we're busy turning him into something useful.
August 4, 2025 at 9:04 AM
Meet Rick, the Unitree G1 we're hacking in the lab. When he's not sneaking out for K-TV, we're busy turning him into something useful.
Introducing Lucy: 1.7B model that Google for you
It's an agentic‑search model that can even run on your phone.
- Agentic search on tap
- Lucy calls tools (
Models
- Lucy 1.7B 40k: huggingface.co/Menlo/Lucy
- Lucy 1.7B 128K: huggingface.co/Menlo/Lucy-...
It's an agentic‑search model that can even run on your phone.
- Agentic search on tap
- Lucy calls tools (
Models
- Lucy 1.7B 40k: huggingface.co/Menlo/Lucy
- Lucy 1.7B 128K: huggingface.co/Menlo/Lucy-...
July 22, 2025 at 9:29 AM
Introducing Lucy: 1.7B model that Google for you
It's an agentic‑search model that can even run on your phone.
- Agentic search on tap
- Lucy calls tools (
Models
- Lucy 1.7B 40k: huggingface.co/Menlo/Lucy
- Lucy 1.7B 128K: huggingface.co/Menlo/Lucy-...
It's an agentic‑search model that can even run on your phone.
- Agentic search on tap
- Lucy calls tools (
Models
- Lucy 1.7B 40k: huggingface.co/Menlo/Lucy
- Lucy 1.7B 128K: huggingface.co/Menlo/Lucy-...
Check @huggingface's homepage, Jan‑nano is trending.
June 20, 2025 at 5:14 AM
Check @huggingface's homepage, Jan‑nano is trending.
Meet Jan-nano, a 4B model that outscores DeepSeek-v3-671B using MCP.
It's built on Qwen3-4B with DAPO fine-tuning, it handles:
- real-time web search
- deep research
Model + GGUF: huggingface.co/collections...
Full technical report will be published shortly.
It's built on Qwen3-4B with DAPO fine-tuning, it handles:
- real-time web search
- deep research
Model + GGUF: huggingface.co/collections...
Full technical report will be published shortly.
June 17, 2025 at 3:04 AM
Meet Jan-nano, a 4B model that outscores DeepSeek-v3-671B using MCP.
It's built on Qwen3-4B with DAPO fine-tuning, it handles:
- real-time web search
- deep research
Model + GGUF: huggingface.co/collections...
Full technical report will be published shortly.
It's built on Qwen3-4B with DAPO fine-tuning, it handles:
- real-time web search
- deep research
Model + GGUF: huggingface.co/collections...
Full technical report will be published shortly.
Singapore AI Showcase is happening tomorrow!
We're bringing researchers, builders, and founders together for an evening of fast-paced lightning talks.
📅 Tuesday, 17 June 2025
🕖 7:00 – 8:30 PM (doors 6:30)
📍 National Library Board, Level 7 — Launch Programme Room
✅ RSVP: lu.ma/wogtowwn
We're bringing researchers, builders, and founders together for an evening of fast-paced lightning talks.
📅 Tuesday, 17 June 2025
🕖 7:00 – 8:30 PM (doors 6:30)
📍 National Library Board, Level 7 — Launch Programme Room
✅ RSVP: lu.ma/wogtowwn
June 16, 2025 at 11:20 AM
Singapore AI Showcase is happening tomorrow!
We're bringing researchers, builders, and founders together for an evening of fast-paced lightning talks.
📅 Tuesday, 17 June 2025
🕖 7:00 – 8:30 PM (doors 6:30)
📍 National Library Board, Level 7 — Launch Programme Room
✅ RSVP: lu.ma/wogtowwn
We're bringing researchers, builders, and founders together for an evening of fast-paced lightning talks.
📅 Tuesday, 17 June 2025
🕖 7:00 – 8:30 PM (doors 6:30)
📍 National Library Board, Level 7 — Launch Programme Room
✅ RSVP: lu.ma/wogtowwn
We're at Echelon Singapore 2025 this week. Come say hi at booth M48 if you’re around. We're showing the agent that gets things done and a humanoid robot.
June 10, 2025 at 3:40 AM
We're at Echelon Singapore 2025 this week. Come say hi at booth M48 if you’re around. We're showing the agent that gets things done and a humanoid robot.
AI Meetup HCMC is tomorrow at VNG Campus.
Lightning talks + networking with people building in AI.
- District 7, HCMC
- 9:00 AM – 12:00 PM
- Mostly in Vietnamese
- RSVP required: lu.ma/7q2ihbj6
Lightning talks + networking with people building in AI.
- District 7, HCMC
- 9:00 AM – 12:00 PM
- Mostly in Vietnamese
- RSVP required: lu.ma/7q2ihbj6
May 30, 2025 at 12:33 PM
AI Meetup HCMC is tomorrow at VNG Campus.
Lightning talks + networking with people building in AI.
- District 7, HCMC
- 9:00 AM – 12:00 PM
- Mostly in Vietnamese
- RSVP required: lu.ma/7q2ihbj6
Lightning talks + networking with people building in AI.
- District 7, HCMC
- 9:00 AM – 12:00 PM
- Mostly in Vietnamese
- RSVP required: lu.ma/7q2ihbj6
Speechless: instruction tuning for speech, without speech.
We're releasing new models, datasets, and a method for training speech instruction models without using any audio data. No waveforms, no TTS.
Links below.
We're releasing new models, datasets, and a method for training speech instruction models without using any audio data. No waveforms, no TTS.
Links below.
May 26, 2025 at 4:58 AM
Speechless: instruction tuning for speech, without speech.
We're releasing new models, datasets, and a method for training speech instruction models without using any audio data. No waveforms, no TTS.
Links below.
We're releasing new models, datasets, and a method for training speech instruction models without using any audio data. No waveforms, no TTS.
Links below.
Come say hi at COMPUTEX 2025!
Menlo is an open R&D lab building the brain for robots. We're around talking LLMs, hardware, robotics, and open source.
Daniel Ong - CEO
Hien To Minh - DevOps
Emre Can Kartal - Marketing
#COMPUTEX2025
Menlo is an open R&D lab building the brain for robots. We're around talking LLMs, hardware, robotics, and open source.
Daniel Ong - CEO
Hien To Minh - DevOps
Emre Can Kartal - Marketing
#COMPUTEX2025
May 16, 2025 at 5:59 AM
Come say hi at COMPUTEX 2025!
Menlo is an open R&D lab building the brain for robots. We're around talking LLMs, hardware, robotics, and open source.
Daniel Ong - CEO
Hien To Minh - DevOps
Emre Can Kartal - Marketing
#COMPUTEX2025
Menlo is an open R&D lab building the brain for robots. We're around talking LLMs, hardware, robotics, and open source.
Daniel Ong - CEO
Hien To Minh - DevOps
Emre Can Kartal - Marketing
#COMPUTEX2025
April 29, 2025 at 4:49 PM
Xin chào! We're hosting our AI Meetup HCMC on Saturday, May 31 at VNG Campus!
This meetup is for anyone building, researching, or learning AI; engineers, founders, students, and researchers to share real work, meet peers, and learn together.
RSVP here: lu.ma/7q2ihbj6 (required to attend)
This meetup is for anyone building, researching, or learning AI; engineers, founders, students, and researchers to share real work, meet peers, and learn together.
RSVP here: lu.ma/7q2ihbj6 (required to attend)
April 29, 2025 at 6:59 AM
Xin chào! We're hosting our AI Meetup HCMC on Saturday, May 31 at VNG Campus!
This meetup is for anyone building, researching, or learning AI; engineers, founders, students, and researchers to share real work, meet peers, and learn together.
RSVP here: lu.ma/7q2ihbj6 (required to attend)
This meetup is for anyone building, researching, or learning AI; engineers, founders, students, and researchers to share real work, meet peers, and learn together.
RSVP here: lu.ma/7q2ihbj6 (required to attend)
ReZero is a reinforcement learning framework that doesn't give up on web search.
It keeps refining its answer until it gets it right. Trained to search, not guess.
Built with synthetic search engines that force it to persist.
It keeps refining its answer until it gets it right. Trained to search, not guess.
Built with synthetic search engines that force it to persist.
April 21, 2025 at 6:13 AM
ReZero is a reinforcement learning framework that doesn't give up on web search.
It keeps refining its answer until it gets it right. Trained to search, not guess.
Built with synthetic search engines that force it to persist.
It keeps refining its answer until it gets it right. Trained to search, not guess.
Built with synthetic search engines that force it to persist.
Models can act better if we describe the world in terms they understand. AlphaSpace gives them that language.
Read more in our blog: menlo.ai/blog/alpha-...
- Code: github.com/menloresear...
- Model: huggingface.co/Menlo/Alpha...
Read more in our blog: menlo.ai/blog/alpha-...
- Code: github.com/menloresear...
- Model: huggingface.co/Menlo/Alpha...
April 10, 2025 at 7:32 AM
Models can act better if we describe the world in terms they understand. AlphaSpace gives them that language.
Read more in our blog: menlo.ai/blog/alpha-...
- Code: github.com/menloresear...
- Model: huggingface.co/Menlo/Alpha...
Read more in our blog: menlo.ai/blog/alpha-...
- Code: github.com/menloresear...
- Model: huggingface.co/Menlo/Alpha...
This shift, structure over perception, was enough to beat the big names.
On EmbodiedBench:
- 10/12 picking tasks
- 6/12 stacking tasks
AlphaSpace offers a better grounding.
On EmbodiedBench:
- 10/12 picking tasks
- 6/12 stacking tasks
AlphaSpace offers a better grounding.
April 10, 2025 at 7:32 AM
This shift, structure over perception, was enough to beat the big names.
On EmbodiedBench:
- 10/12 picking tasks
- 6/12 stacking tasks
AlphaSpace offers a better grounding.
On EmbodiedBench:
- 10/12 picking tasks
- 6/12 stacking tasks
AlphaSpace offers a better grounding.
AlphaSpace turns 3D scenes into structured tokens. Object shape, color, position, height - captured as clean text.
It uses a two-level grid to represent space:
- Coarse: locate objects in broad zones
- Fine: pin down their precise position
The model sees: <|0-1|><|local-3-2|><|red|><|cube|>
It uses a two-level grid to represent space:
- Coarse: locate objects in broad zones
- Fine: pin down their precise position
The model sees: <|0-1|><|local-3-2|><|red|><|cube|>
April 10, 2025 at 7:32 AM
AlphaSpace turns 3D scenes into structured tokens. Object shape, color, position, height - captured as clean text.
It uses a two-level grid to represent space:
- Coarse: locate objects in broad zones
- Fine: pin down their precise position
The model sees: <|0-1|><|local-3-2|><|red|><|cube|>
It uses a two-level grid to represent space:
- Coarse: locate objects in broad zones
- Fine: pin down their precise position
The model sees: <|0-1|><|local-3-2|><|red|><|cube|>