TII Releases Falcon-ASR: 1.6B Arabic Speech Model Hits 20.92% WER
The Technology Innovation Institute (TII) in Abu Dhabi has released Falcon-ASR, a 1.6B parameter Arabic speech recognition model.
Proof1 independent outlet
Key points
- In TII’s evaluation, Falcon-ASR posts a 20.92% average word error rate (WER) across 6 Arabic test sets.
- Performance: Lowest Arabic and Emirati WER in TII’s own comparisons, with solid English results.
- Falcon-ASR is a speech-to-text model from TII that converts spoken Arabic and 4 other languages into text.
- How does Falcon-ASR handle Arabic dialects?
Sources (1)
- [1]TII Releases Falcon-ASR: 1.6B Arabic Speech Model Hits 20.92% WERMarkTechPost · Oct 12, 03:16 AM
The Technology Innovation Institute (TII) in Abu Dhabi has released Falcon-ASR, a 1.6B parameter Arabic speech recognition model.
In TII’s evaluation, Falcon-ASR posts a 20.92% average word error rate (WER) across 6 Arabic test sets.
Extractive summary: sentences quoted from the sources.
Hype check
Independently evaluated1 independent look since
- Evaluation · Oct 12, 2026PatTransBench: an ipc-stratified, multi-dimensional benchmark for patent machine translation evaluation
- Follow-up · Oct 12, 2026What are you doing for a reliable, well designed local model based AI search?
- Follow-up · Oct 12, 2026Hardware-Friendly Token-Group Activation Quantization for Low-Bit Mamba Super-Resolution
- Follow-up · Oct 12, 2026Fast Quantized Johnson$-$Lindenstrauss Transforms
- Follow-up · Oct 12, 2026QLoRA on ExLlamaV3 Quants - Train @ 2-8bpw
- Follow-up · Oct 12, 2026Strong yet Controllable LLM Abstention without Learning to Abstain
Before this
- Oct 11, 2026Putting Earth AI to work: Geospatial agents for our planet's biggest challenges
- Oct 11, 2026AesCode: Aesthetic Code Generation with Decoupled Cross-Modal Rewards
- Oct 10, 2026vllm-sr/d3
- Oct 10, 2026SAGE: Pushing 2-Bit Vector-Quantized LLMs toward Real-World Usability
- Oct 10, 2026KV4-RL: Unlocking Faster Asynchronous Rollouts for RLVR with INT4 KV Cache Quantization
- Oct 9, 2026Qwen/Qwen-Image-2.1-Turbo