Overview
Falcon 3 from Abu Dhabi’s TII was trained on 14 trillion tokens and focuses on science, math and code performance in small sizes.
Highlights
- 1B, 3B, 7B and 10B sizes
- 32K-token context window
- Trained on 14 trillion tokens
- Permissive Falcon license
Typical uses
Local assistants and fine-tuning on modest hardware.
Things to consider
- Smaller ecosystem than Llama or Qwen.
- Download sizes refer to the default quantized versions in the Ollama library; full-precision weights are larger.
Specifications are taken from the model card on Hugging Face, the Ollama library and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.