Lucas Atkins
@LucasAtkins7
cto @arcee_ai
We are hiring several additional cloud engineers to help us continue building the most versatile SLM and tooling deployment stack in the industry. This is a high priority for us, and we will fast-track the hiring process. DM me with questions or apply below.
I love when doordash hits you with the "Uh you're not near that address, stupid"
Good times yesterday at the MCP Conference in London (lu.ma/mcpconference). To cut through the hype, I demoed how to build single- and multi-agent MCP applications from scratch, using only open-source tools from @huggingface (smolagents, Gradio) and @arcee_ai models…
I can confirm this model is rather amazing
>>> Qwen3-Coder is here! ✅ We’re releasing Qwen3-Coder-480B-A35B-Instruct, our most powerful open agentic code model to date. This 480B-parameter Mixture-of-Experts model (35B active) natively supports 256K context and scales to 1M context with extrapolation. It achieves…
Whatever RL magic they cooked up to facilitate this is insane, well done all around.
Bye Qwen3-235B-A22B, hello Qwen3-235B-A22B-2507! After talking with the community and thinking it through, we decided to stop using hybrid thinking mode. Instead, we’ll train Instruct and Thinking models separately so we can get the best quality possible. Today, we’re releasing…
We've just release 100+ intermediate checkpoints and our training logs from SmolLM3-3B training. We hope this can be useful to the researcher working on mech interpret, training dynamics, RL and other topics :) Training logs: -> Usual training loss (the gap in the loss are due…
Soon
Apple Intelligence Foundation Language Models Tech Report 2025 sorry apple AFM is reserved for @arcee_ai
Imagine if Mcfly had twitter. He’d be like “@doc is this true?”
Imagine if sheep had twitter. They’d be like “@flock is this true?”
amortization you can always* improve the perf of an agent by scaling up inference, with logarithmic returns if you want to use an agent once and don't care about cost, just use an API if you want to deploy something that will run many similar workloads and cost matters, you…