Sarvam has launched the Sarvam Token Factory, an inference platform hosted in India that provides access to eight open-weight models via a single OpenAI-compatible API. The service utilizes NVIDIA B200 clusters and aims to offer low-latency inference for Indian developers by keeping traffic within the country. Sarvam reports significant performance gains, including faster first-token response times and higher throughput compared to competitors like InferenceX, particularly for its own Sarvam 105B model. AI
IMPACT Provides localized, low-latency inference options for developers in India, potentially reducing costs and improving performance for AI applications.
RANK_REASON Product launch of an inference platform by a specific company, not a frontier model release from a major lab.
- Gemma 4 31B
- GLM 5.3
- India
- InferenceX
- Kimi K3.1 Pro
- Mumbai
- Nvidia B200
- Qwen 3.8-27B
- Sarvam
- Sarvam Token Factory
- vLLM
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →