Intron Unveils Sahara v2.5 AI Model Built for African Code-Switched Speech

By Adedapo Adesanya

Intron, an Africa-centric voice technology platform, has launched Sahara v2.5, the next phase for African voice AI that enables computers to understand one of the continent’s most common ways of communicating: mixing languages within the same sentence and conversation.

Across Africa, multilingual conversations are a routine part of daily life, with people often switching between languages to communicate more effectively and express identity and context. A doctor may explain a diagnosis in English before reassuring a patient in Swahili, while a bank customer could discuss a loan in Yoruba and complete the conversation in English or Nigerian Pidgin.

However, many artificial intelligence systems struggle to accurately process such code-switching conversations, often missing critical segments or requiring users to stick to one language. Sahara v2.5 is designed to address this challenge by improving AI’s ability to understand and respond to multilingual conversations across African languages.

The new release introduces bilingual language mixing (code-switching) speech recognition to 12 African languages, including Zulu, Hausa, Swahili, and Luganda. Intron’s published benchmarks show Sahara outperforming giants like Gemini, ElevenLabs and Meta on all 12 languages tested.

Intron also introduces the world’s first African trilingual speech-recognition model that supports switching between Kinyarwanda, English and French, using proprietary algorithms and technology for which the company has filed US patents.

The firm said it has also added fluent text-to-speech and voice-agent capabilities with language-mixing across the 13 languages, outperforming ElevenLabs and Gemini in 11 of 13 languages tested, enabling businesses to build voice experiences, voice bots, and voice agents that sound natural to local users. New speech-recognition support for Nupe, Kanuri, Nigerian Fulfulde, Tigrinya, Kikuyu, Dholuo, and Somali expands Sahara’s African-language coverage to 31 languages, making it one of the continent’s most comprehensive voice AI platforms.

For enterprises, the Sahara v2.5 breakthrough will reduce the need for customers to change how they speak to be understood. Contact centres, hospitals, financial institutions, governments and WhatsApp voice assistants can now follow conversations as they naturally move between local languages and English/French, making voice AI more accurate, more inclusive and ultimately more useful for the people it serves.

With offline models deployed on Nvidia hardware at PAMO Clinics in Port Harcourt, Nigeria, through a global donor-funded project, Intron brings private and sovereign AI closer to enterprises, governments, and institutions where regulatory or connectivity constraints limit access to best-in-class technology.

Since raising $1.6 million in pre-seed funding in 2024, Intron has expanded its training data to over 150,000 hours of African-language audio from over 53,000 speakers, covering 64 languages and more than 500 accents. Code-switching support includes Afrikaans-English, Akan-English, Amharic-English, Hausa-English, Igbo-English, Kinyarwanda-English, Kinyarwanda-French, Luganda-English, Nigerian Pidgin English, Swahili-English, Wolof-French, Yoruba-English, Zulu-English.

Sahara v2.5 also introduces new streaming speech-recognition and text-to-speech capabilities to the API supporting live captions, real-time applications and real-time speech generation. Several Improvements to latency, concurrency and reliability support high-volume applications across chatbots, contact centres, medical documentation, financial services, legal workflows, and agricultural or climate advisory services.

Sahara already supports organisations across healthcare, legal, financial services and contact centres. Existing deployments include Branch to recover more than N1.2 million in delinquent loans in one week, with record after-hours and weekend repayments, outperforming human agents on delinquent loans over 356 days; Audere Africa to allow South African youths to express themselves freely instead of being confined to text chat; the Ogun State Judiciary to automate transcription across all 18 high courts in the state; and also saw a typical 14-minute Swahili-English doctor-patient consultation in Nairobi converted into a structured clinical note in less than 30 seconds, taking the documentation burden off overworked physicians so they can focus more on patient care.

Speaking on the latest development, Ms Adanne Anene, Head of Product Africa at Branch International, said: “Collaborating with Intron to build a Branch-aligned collections bot was a rewarding experience; customers engaged naturally even after hours and on weekends. Conversations were human and effective at delivering real payments on delinquent loans, sometimes over two years old. Adding local language and language-mixing support in this v2.5 release helps us reach even more customers.”

Sahara also supports production and research deployments for over 40 enterprise customers across 6 countries, including Nigeria, Kenya, South Africa, Uganda, Rwanda and Ghana.

Intron said Sahara v2.5 recorded a lower word error rate of 34.3% across 12 African languages, compared with 53.8% for Gemini 3.6, while separate tests found it outperformed Gemini and Meta’s Omnilingual model in five of seven Nigerian languages.

“Code-switching was one of the biggest problems that consistently came up for clients deploying real-world voice AI. The content loss with most models increases post-editing time, and propagates errors and omissions to downstream summaries, call logs, court records, and clinical notes,” said Mr Tobi Olatunji, CEO of Intron, adding that, “Africa needs AI built for how Africans really speak. People should not have to translate themselves for a machine, flatten their accent, avoid local expressions or repeat only the English part of what they said. Voice AI should work with the way people already speak.”

Along with the launch of Sahara v2.5, Intron has published its 2026 Africa Voice AI Report. The report challenges the assumption that collecting African language data is the only barrier to reliable voice AI, arguing that research capacity, orchestration, and implementation expertise weigh as heavily as hours of audio. Ambient medical scribes make the case– although widely adopted across the US and Europe, they have struggled in African clinics where consultations move between languages. Access the report here.

At the 2026 Deep Learning Indaba in Lagos, Africa’s largest AI gathering, the Sahara CodeSwitch Hackathon brought together over 120 teams from 21 countries to build fintech, healthtech, agritech, and edutech apps that support language-mixing on Sahara. Several teams compared Sahara with several open, closed, and commercial models, including Gemini, AssemblyAI, and Whisper; the results indicated that Sahara performed significantly better even though it was slightly slower.

The competition continues through September, with a dedicated Masterclass hosted by Nvidia on August 28th. Registration remains open, with teams competing for the $10,000 prize pool.