FrontierAugust 14, 2026via MarkTechPost
Meet Needle 2: An Open 45M-Parameter Tool-Calling Model That Ships as a 14MB Binary and Runs a Full Session in 28MB of RAM
Why it matters
A new open-weight model challenges the assumption that capable AI requires scale and silicon. Needle 2's efficiency (sub-30MB RAM, no accelerators) matters for practitioners building on-device, edge, and resource-constrained deployments — and signals a shift in how frontier labs think about capability-per-watt.
Key signals
- Needle 2: 45M parameters
- Binary size: 14MB
- Session RAM footprint: ~28MB
- No GPU or NPU required
- Leads Seal-Tools benchmark splits
- Open-weight release
- Vendor: Cactus Compute
- Use cases: tool calling, device use, structured extraction
The hook
45M parameters. 14MB binary. Needle 2 runs a full AI session on hardware with zero GPUs—and leads tool-calling benchmarks doing it.
Cactus Compute released Needle 2, an open 45M-parameter model for tool calling, device use, and structured extraction. The full model is a single 14MB binary that runs a session in about 28MB of RAM. It leads both Seal-Tools splits while targeting hardware with no GPU and no NPU.
The post Meet Needl…