Menlo Research's banner

Menlo Research

@menloresearch • 4,807 subscribers

We work with forward looking organizations to build their robot labor force of tomorrow. Get in touch: https://t.co/DQeXXgnvAr

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

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: To run it locally: - Install Jan Beta: - Download Jan-nano in Jan Hub - Settings -> MCP, enable MCP and add your Serper API key for web tools Full technical report will be published shortly.

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: To run it locally: - Install Jan Beta: - Download Jan-nano in Jan Hub - Settings -> MCP, enable MCP and add your Serper API key for web tools Full technical report will be published shortly.

56,600 views • 1 year ago

🍓 Ichigo-llama3.1: Local Real-Time Voice AI We bring 2 key improvements to Ichigo: - It can talk back (Yes!) - It recognizes when it can't comprehend input You can now run this little strawberry on your device! Demo on a single NVIDIA 3090 GPU. 1/10

🍓 Ichigo-llama3.1: Local Real-Time Voice AI We bring 2 key improvements to Ichigo: - It can talk back (Yes!) - It recognizes when it can't comprehend input You can now run this little strawberry on your device! Demo on a single NVIDIA 3090 GPU. 1/10

67,175 views • 1 year ago

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 ( ‑aware) - Fits in your pocket - runs on CPU or mobile Under the hood: - Built on Qwen's Qwen3‑1.7B - Smooth multi‑category rewards replace brittle if‑else scoring - Task‑vector RLVR optimizes the "thinking" tag for targeted search moves. Benchmarks: - SimpleQA + MCP = 78.3 - Close to Jan‑Nano-4B (80.7) Run locally: - Demo uses vLLM in Jan - You can spin it up with Georgi Gerganov's llama.cpp or vLLM Models on Hugging Face: - Lucy 1.7B 40k: - Lucy 1.7B 128K:

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 ( ‑aware) - Fits in your pocket - runs on CPU or mobile Under the hood: - Built on Qwen's Qwen3‑1.7B - Smooth multi‑category rewards replace brittle if‑else scoring - Task‑vector RLVR optimizes the "thinking" tag for targeted search moves. Benchmarks: - SimpleQA + MCP = 78.3 - Close to Jan‑Nano-4B (80.7) Run locally: - Demo uses vLLM in Jan - You can spin it up with Georgi Gerganov's llama.cpp or vLLM Models on Hugging Face: - Lucy 1.7B 40k: - Lucy 1.7B 128K:

20,205 views • 11 months ago

ReZero: A small model that learns to search - it never gives up 🔥 ReZero trains with synthetic search engines that force the model to retry, refine, and persist until it finds a better answer (never give up 💪). It's built on Meta's Llama 3.2B. Instead of optimizing for recall or speed, we train the model to retry when it's wrong - using reinforcement learning to build persistence into the search process. - Model: - Code: Thanks to AI at Meta for the Llama 3.2B base, Unsloth AI for AutoDidact (the framework we built on), and Colin Kealty for quantizing the model!

ReZero: A small model that learns to search - it never gives up 🔥 ReZero trains with synthetic search engines that force the model to retry, refine, and persist until it finds a better answer (never give up 💪). It's built on Meta's Llama 3.2B. Instead of optimizing for recall or speed, we train the model to retry when it's wrong - using reinforcement learning to build persistence into the search process. - Model: - Code: Thanks to AI at Meta for the Llama 3.2B base, Unsloth AI for AutoDidact (the framework we built on), and Colin Kealty for quantizing the model!

13,948 views • 1 year ago

No more content to load