Sarvam releases open-weight Sarvam-105B-1711 for low-cost agents
Sarvam AI has released Sarvam-105B-1711, a new open-weight checkpoint of its 105-billion-parameter model built for high-volume work such as customer support, voice agents and agentic tasks, the company says. Across eight conversational agent benchmarks, Sarvam says the model averages 73.9 at a blended $0.15 per million tokens, against 74.0 for Gemini 3.5 Flash-Lite at $0.85 and 73.0 for Gemma 4 31B at $0.54, on its own evaluation setup.
The biggest gains since the March version are in instruction following, which rose from 32.5 to 82.5 on IFBench, tool use in customer-service conversations, from 46.2 to 85.5 on Tau2, and Indian languages, from 30.1 to 79.1 on Sarvam's Indic suite. The company says it solved 69.5% of Tau2 tasks in all four of four tries, the most consistent of the models it tested. Training in 8-bit precision on Nvidia Blackwell roughly doubled pretraining throughput to 10,900 tokens per second per GPU, and long-context training on 512 GPUs extended the context window to 160K tokens.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like Daily AI Brief.
Also, consider following us on social media:
Daily AI Brief
Daily report covering major AI developments and industry news, with both top stories and complete market updates
Industry analysis
2025 Global Business Services Agenda: Gen AI Takes Center Stage
This industry analysis by The Hackett Group explores the transformative impact of generative artificial intelligence (Gen AI) on global business services (GBS) in 2025. The study highlights the shift from exploration to acceleration of Gen AI initiatives, with 89% of executives advancing these projects to improve customer satisfaction, innovate products, and reduce costs. The report also discusses the challenges and strategies for successful Gen AI adoption, emphasizing the need for a technology-enabled operating model and the importance of reskilling the workforce.
Read moreYou may also like
Sarvam AI opens Token Factory inference service on Indian GPU clusters
Salesforce releases SOV-Transcribe-1, an open-weight speech-to-text model
Sema4.ai Releases Agent Platform 3.0 With Reusable Logic
Shanghai AI Lab releases Sage S2.1 Flash, a 321B open scientific multimodal model
Agnes AI releases open-weight Agnes 2.5 Pro Theta, a 188B multimodal model
Daily AI Brief: the AI news that matters, in your inbox.