Researchers wanting a small reasoning model to run locally get free weights, though the Falcon licence is not a standard open-source one.
Kenning's take
What it does
Falcon H1R 7B is a 7 billion parameter language model built on the hybrid Falcon-H1 base and trained further for step-by-step reasoning in maths, coding and science. TII reports that it matches or beats reasoning models two to seven times its size on several benchmarks while using fewer tokens. Full and quantized GGUF weights are free to download on Hugging Face under the Falcon LLM licence.
- 7B hybrid architecture
- Reasoning traces up to 48k tokens
- Trained with supervised fine-tuning and GRPO reinforcement learning
- Test-time scaling with DeepConf
- Full and GGUF quantized weights
Who it suits. Developers and researchers who want a small reasoning model to run locally.
Key facts
- Category
- AI models
- Made by
- Technology Innovation Institute (TII)
- Based in
- United Arab Emirates (Abu Dhabi)
- Released
- 5 January 2026
- How to get it
- Download weights from Hugging Face
- Works on
- open weights, Hugging Face
- In the UK
- Yes
- Free plan
- open weights, Falcon LLM licence
FAQ
When was Falcon H1R 7B released?
TII introduced it on 5 January 2026.
Is Falcon H1R 7B free?
Yes, the weights are free on Hugging Face under the Falcon LLM licence.
Who made Falcon H1R 7B?
The Technology Innovation Institute in Abu Dhabi.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.