Microsoft AI's real-time speech-to-text model for voice agents, transcribing 60 languages with continuous language detection
TrustList summary
A raw, pretrained base model rather than a finished assistant: that is how the Hugging Face card introduces Falcon 2 11B, a causal decoder-only model with 11 billion parameters from the Technology Innovation Institute (TII).
A foundation for further tuning
The card recommends fine-tuning the model for most real uses. Its direct uses are research on large language models and, after specialisation, tasks such as summarisation, text generation and chatbots. Production use without a proper assessment of risks and mitigations is listed as out of scope. The licence is the TII Falcon License 2.0, which TII describes as permissive and based on Apache 2.0, with an acceptable use policy that promotes responsible use of AI.
Languages and stated limits
It was trained mostly on English but also on German, Spanish, French, Italian, Portuguese, Polish, Dutch, Romanian, Czech and Swedish, and the card says it will not generalise well to other languages. Because the training data is a large web corpus, it carries the stereotypes and biases commonly found online, so TII advises guardrails for any production deployment.
Training in four stages
Training used more than 5,000 billion tokens of RefinedWeb, a filtered and deduplicated web dataset enhanced with curated corpora. The card lists English and European web data, technical data, code and conversational data from public sources. The first three stages raised the context length from 2,048 to 4,096 and then 8,192 tokens, and a final stage used only high-quality data. Most training ran on 1,024 A100 40GB GPUs using 3D parallelism with ZeRO and Flash-Attention 2, in bfloat16 precision with the AdamW optimiser. An arXiv technical report and a Hugging Face blog post give more detail. Running it with the transformers library requires PyTorch 2.0.
Summary written by TrustList from the business's own published information, Nov 2026.
Benchmarks & AI stats
| Benchmark | Official | Community avg |
|---|---|---|
| No benchmark scores yet. Be the first to add one. | ||
“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).
Sign in to add a benchmark score for this model.
Falcon 2 11B launches
Request a demo or quote from Falcon 2 11B
Reviews
Write the first review of Falcon 2 11B
Used it? Your experience helps other buyers decide.
Write a reviewQuestions & answers
No questions yet. Be the first to ask about Falcon 2 11B.
Falcon 2 11B alternatives
Similar listings buyers compare against this one.
Liquid AI's open-weight decision model, giving yes or no, choice or score answers from text and images
Cantina Security's first open-weight security model, a post-trained GLM-5.3-Flash for vulnerability research and exploit checks
Aleph Alpha's open-weight English-German mixture-of-experts model for self-hosting, released under Apache 2.0 in October 2026
Open-source 2B decision model from AWS's Strands Agents project that picks options and returns scores with a reliability figure
Perplexity's open-weight decision model that picks among given options for routing and checks in AI systems
Google's open multimodal embedding model that maps text, images, audio and video into one vector space, under Apache 2.0
JetBrains' open 12B mixture-of-experts coding model with 2.5B parameters active per token under Apache 2.0
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- Write a reviewHelp other buyers decide
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- ArticlesGuides, comparisons and research from the editorial desk and the community.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.