Nigerian voice AI startup Intron has launched Sahara v2.5, a significant update to its African voice technology platform that enables artificial intelligence to understand code-switching, the
Nigerian voice AI startup Intron has launched Sahara v2.5, a significant update to its African voice technology platform that enables artificial intelligence to understand code-switching, the widespread African practice of moving between languages within the same sentence or conversation.
The announcement was made on Tuesday, August 26, 2026, at the Deep Learning Indaba in Lagos, Africa’s largest AI gathering. The update addresses a gap that has long frustrated enterprises deploying voice AI across the continent. A doctor explaining a diagnosis in English before switching to Swahili to reassure a patient. A bank customer discussing a loan in Yoruba before finishing the sentence in English. A court proceeding that moves between languages without warning.
Most AI systems struggle to follow these conversations, dropping critical segments and forcing users to either choose a single language or repeat themselves. Sahara v2.5 is built to follow the conversation as it naturally happens.
The new release introduces bilingual language mixing speech recognition to 12 African languages, including Zulu, Hausa, Swahili and Luganda. It also introduces the world’s first African trilingual speech recognition model, supporting switching between Kinyarwanda, English and French, using proprietary algorithms for which Intron has filed US patents.
Total African language coverage now stands at 31, with new support added for Nupe, Kanuri, Nigerian Fulfulde, Tigrinya, Kikuyu, Dholuo and Somali.
Intron’s published benchmarks show Sahara v2.5 outperforming global models including Gemini, ElevenLabs and Meta across all 12 languages tested. Measured by word error rate, where a lower score means fewer errors, Sahara achieved an average of 34.3% across 12 languages, compared with 53.8% for Gemini 3.6 — a 36% relative reduction in errors.
Separate benchmarking conducted through Gooey.ai for the Gates Foundation and CLEAR Global found Sahara performed better on five of seven Nigerian languages evaluated, including against Gemini and Meta’s Omnilingual model.
“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,” said Tobi Olatunji, CEO of Intron.
Read also: Google offers Nigerian students a free AI Plus subscription for 12 months
Intron puts Sahara to work across African industriesSahara v2.5 is not a research announcement. It is already deployed across enterprise customers in healthcare, legal, financial services and contact centres across Nigeria, Kenya, South Africa, Uganda, Rwanda and Ghana.
At Branch International, Sahara-powered collections agents recovered more than ₦1.2 million in delinquent loans in one week, with record after-hours and weekend repayments, outperforming human agents on loans delinquent for over 356 days. Adanne Anene, Head of Product Africa at Branch International, described the experience as conversations that felt human and effective even on loans more than two years old.
At the Ogun State Judiciary, Sahara has expanded from a single pilot court to nine courts over the past year, with plans to automate transcription across all 18 high courts in the state. Proceedings that previously took more than four hours now finish in half the time.
In Nairobi, a 14-minute Swahili-English doctor-patient consultation is converted into a structured clinical note in less than 30 seconds, reducing documentation burden on overworked physicians. In South Africa, young people can now speak naturally with Audere Africa’s Self-Cav Reproductive Health WhatsApp chatbot using voice, rather than being confined to text.
Intron has also deployed offline models on Nvidia hardware at PAMO Clinics in Port Harcourt through a global donor-funded project, bringing private and sovereign AI to institutions where regulatory or connectivity constraints limit access to technology.
Sahara v2.5 also introduces new streaming speech recognition and text-to-speech capabilities to its API, supporting live captions, real-time applications and real-time speech generation, with improvements to latency, concurrency and reliability for high-volume applications.
Alongside the launch, Intron published its 2026 Africa Voice AI Report, which challenges the assumption that collecting African language data is the only barrier to reliable voice AI. The report argues that research capacity, orchestration and implementation expertise weigh as heavily as hours of audio. 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 more than 53,000 speakers, covering 64 languages and more than 500 accents.
The company currently supports over 40 enterprise customers across six countries. The Sahara CodeSwitch Hackathon, held at the Deep Learning Indaba, brought together over 120 teams from 21 countries to build fintech, healthtech, agritech and edutech applications on the platform. The competition continues through September, with a Masterclass hosted by Nvidia on August 28th and a $10,000 prize pool for top-performing teams.
