AI API pricing in India — INR per million tokens
DeepInfra Open models
₹8/M tokens · 280 ms from Mumbai · GST: extra · Cheapest open-weight serve
Hyperbolic deepseek-v4
₹18/M tokens · 240 ms from Mumbai · GST: extra · Cheapest open-weight DeepSeek serve
Krutrim Krutrim-1
₹35/M tokens · 90 ms from Mumbai · GST: included · Indic + cost, 22 languages
Sarvam Sarvam-1
₹40/M tokens · 80 ms from Mumbai · GST: included · Indic-native, INR billing, Mumbai latency
xAI SpaceXAI: Grok 4.3
₹104/M tokens · 280 ms from Mumbai · GST: extra · Reasoning + agentic instruction following
Mistral AI Mistral Large
₹167/M tokens · 260 ms from Mumbai · GST: extra · European-hosted, strong tool-use
Fireworks Mistral: Mixtral 8x22B Instruct
₹167/M tokens · 200 ms from Mumbai · GST: extra · Fast inference, model variety
Perplexity Perplexity: Sonar Pro
₹250/M tokens · 230 ms from Mumbai · GST: extra · Search-grounded responses, citations
Anthropic Anthropic: Claude Sonnet 4.6
₹250/M tokens · 300 ms from Mumbai · GST: extra · Production coding workhorse
Anthropic Anthropic: Claude Opus 4.7
₹418/M tokens · 320 ms from Mumbai · GST: extra · Coding, long-context, agentic workflows
OpenAI OpenAI: GPT-5.5 Pro
₹2,505/M tokens · 280 ms from Mumbai · GST: extra · Frontier reasoning, voice, structured outputs
Cohere command-r-plus
300 ms from Mumbai · GST: extra · Enterprise retrieval + RAG
Bhashini Bhashini NMT
110 ms from Mumbai · GST: included · 22-language Indian govt translation API
Replicate (per-model)
280 ms from Mumbai · GST: extra · Wide model catalogue, on-demand
Amazon Bedrock (per-model)
160 ms from Mumbai · GST: extra · Enterprise AWS-native
Google gemini-3.1-flash
120 ms from Mumbai · GST: included · Best price/perf, multimodal, Indic capable
Google gemini-3.1-pro
150 ms from Mumbai · GST: included · Frontier multimodal, long context
DeepSeek deepseek-v4
260 ms from Mumbai · GST: extra · Million-token context + FP4 economy
Azure OpenAI (per-model)
140 ms from Mumbai · GST: extra · Microsoft-hosted OpenAI for enterprise
OpenRouter (per-model)
280 ms from Mumbai · GST: extra · Unified API across 344+ models
Groq llama-4-70b-instruct
220 ms from Mumbai · GST: extra · Lowest-latency open-weight inference
Together llama-4-70b-instruct
180 ms from Mumbai · GST: extra · Open weights as managed API