# IndieRouter

> IndieRouter is India's sovereign AI inference cloud: an OpenAI-compatible API served exclusively from GPUs in Indian data centres. 11 open-weight models live across chat/completions, speech-to-text, text-to-speech and document OCR; image generation coming soon. DPDP-compliant by architecture; billing in INR with GST invoices.

## Product
- [Models](https://indierouter.ai/models/): Live catalog of open-weight models served from Indian GPUs, with per-model context length, pricing, and model card links.
- [Pricing](https://indierouter.ai/pricing/): India-resident inference pricing in rupees per million tokens, prepaid credit plans, and GST invoicing.
- [Coding plans](https://indierouter.ai/coding-plans/): Monthly subscription plans from ₹499 for OpenAI-compatible coding agents: mint a plan key, included usage value across 5-hour, weekly and monthly rolling windows.
- [Docs](https://indierouter.ai/docs/): Developer docs covering quickstart, authentication, OpenAI-compatible endpoints, streaming, and rate limits.
- [Partners](https://indierouter.ai/partners/): Partner program for agencies, dev shops, systems integrators and hosting providers: referral, white-label reseller, and SI tracks starting at 15%.
- [Enterprise](https://indierouter.ai/enterprise/): Dedicated and Sovereign tiers: one fixed monthly price for unlimited inference on the open-weights model you choose, quoted from concurrency and throughput, plus compliance controls (SSO, BYOK, audit export).

## Blog
- [Seven new models on IndieRouter, served from Noida](https://indierouter.ai/blog/seven-new-models-noida/): GLM-5.3, Gemma 4, Nemotron 3.5 Super VL, Qwen 3.8 27B and 1M-context DeepSeek and GLM Flash, from our new Noida region. Same API, same key.
- [DeepSeek V4.1 Flash 1017 is live on IndieRouter, same id, same price](https://indierouter.ai/blog/deepseek-v4-1-flash-1017/): DeepSeek's October update to V4.1 Flash now serves deepseek-v4.1-flash: stronger on coding and agent benchmarks, shorter outputs, still ₹20 / ₹60 per M tokens.
- [Ling-3.1-flash: a 560B open model built mostly on linear attention](https://indierouter.ai/blog/ling-3-1-flash-open-weights/): inclusionAI released Ling-3.1-flash under MIT: 560B parameters, 25B active, 7 of every 8 layers linear attention. What the benchmarks say and what to check.
- [Part 14: Pushing the serving envelope](https://indierouter.ai/blog/h200-part-14-pushing-the-serving-envelope/): Extend DeepSeek to a million-token ceiling, compare draft lengths and engines, and separate fast benchmark rows from usable serving recipes.
- [Part 15: Two hours and one last check](https://indierouter.ai/blog/h200-part-15-two-hours-and-one-last-check/): A corrected load generator, two long DeepSeek runs, and the HTTP 500 that kept the faster configuration from passing the final gate.
- [After the campaign: the fixes that changed our next run](https://indierouter.ai/blog/h200-afterword-rescues-and-quality-checks/): We revisit the missing GLM baseline, repair DeepSeek loaders, and test native TP4, DSpark and two replicas while keeping the limits of each result visible.
- [Eight H200s: the inference campaign, in 15 parts](https://indierouter.ai/blog/h200-nvl-inference-campaign/): Follow our H200 NVL campaign from the first topology check to a million-token conversation: fifteen engineering stories, measured results and failures.
- [Part 1: Arrival on two H200 NVL islands](https://indierouter.ai/blog/h200-part-01-arrival-two-islands/): Start with the machine: storage, NUMA, NVLink topology, a broken cross-socket bridge, and the IOMMU fix that made it usable.
- [Part 2: Qwen3.8-27B, from smoke test to the 1M wall](https://indierouter.ai/blog/h200-part-02-qwen38-27b-first-model/): Bring up the first model, check quality before speed, compare FP8 and INT4, and work through the attempts to reach one million tokens.
- [Part 3: gpt-oss-120b, two kernels and many users](https://indierouter.ai/blog/h200-part-03-gpt-oss-120b-kernels/): Compare Triton and Marlin on gpt-oss-120b, follow latency as concurrency grows, and keep the failed quality gate attached to the excluded arm.
- [Part 4: Chasing one million tokens on four GPUs](https://indierouter.ai/blog/h200-part-04-million-tokens-one-quad/): Move from fitting the context to finding the needle, reading a fresh million-token document, and sustaining decode at depth.
- [Part 5: The giants: GLM, MiniMax and DeepSeek](https://indierouter.ai/blog/h200-part-05-glm-minimax-deepseek-giants/): Take the largest models across both GPU islands and follow the memory failures, long-prefill costs, and provisional MiniMax results.
- [Part 6: Parallelism on two islands](https://indierouter.ai/blog/h200-part-06-parallelism-two-islands/): Compare TP8, independent replicas and expert parallelism on the actual topology, then investigate the anomaly and the fixes that did not help.
- [Part 7: Everything that broke, and what fixed it](https://indierouter.ai/blog/h200-part-07-failures-and-fixes/): Follow the node and engine failures, successful retries and unresolved paths, including the DeepSeek launches that stopped before benchmarking.
- [Part 8: Tuning and splitting prefill from decode](https://indierouter.ai/blog/h200-part-08-prefill-decode-tuning/): Test a larger prefill budget and a prefill/decode split, keeping the queueing costs and the missing baseline visible in the comparison.
- [Part 9: Speculative decoding, from one user to sixteen](https://indierouter.ai/blog/h200-part-09-speculative-decoding/): Measure when drafting helps chat and code workloads, where concurrency changes the result, and which speculative paths never produced valid numbers.
- [Part 10: Cold starts and the memory below the GPU](https://indierouter.ai/blog/h200-part-10-cold-starts-memory/): Trace weight loading, startup time, host RAM and KV offload, separating a warm page cache from a reproducible cold start.
- [Part 11: What the tokens cost, and why users set the price](https://indierouter.ai/blog/h200-part-11-inference-economics/): Turn the measured throughput into scenario-based costs, state the rental-price assumption, and separate fresh documents from cached conversations.
- [Part 12: Nine models, one node, one million tokens](https://indierouter.ai/blog/h200-part-12-final-head-to-head/): Bring the original campaign together: compare the last turn, the cost of another user, the useful configurations and the unfinished work.
- [Part 13: Where FP8 checkpoints come from](https://indierouter.ai/blog/h200-part-13-fp8-checkpoints/): Inspect tensor headers and scale conventions behind published FP8 names, including the difference between an expert cast and an all-FP8 model.
- [Growing one conversation to a million tokens](https://indierouter.ai/blog/h200-supplement-growing-a-conversation-to-1m/): A supplementary report on cached, growing conversations: per-turn latency, retrieval checks and the difference from a fresh million-token prompt.
- [DeepSeek V4.1 Flash is live on IndieRouter, and cheaper than V4](https://indierouter.ai/blog/deepseek-v4-1-flash-live/): DeepSeek's newest open-weight Flash model, served from our GPUs in India: native vision, tool calling, 262K context, ₹20 / ₹60 per million tokens.
- [DeepSeek opened V4 Flash Vision weights under MIT](https://indierouter.ai/blog/deepseek-v4-flash-vision-open-weights/): Ten days after the API drop, V4 Flash Vision Exp is on Hugging Face under MIT. 305B MoE, 1M context, same token price as text Flash.
- [AMD skipped ROCm 8 and 9. Here is what 10 ships](https://indierouter.ai/blog/amd-rocm-10/): ROCm 10.0 jumps from 7.14 to a TheRock-built Core SDK plus ROCm.AI. The 3.3x inference claim is real lab data, not a free upgrade.
- [Qwen3.8-Flash-Next previews the Qwen4 architecture](https://indierouter.ai/blog/qwen3-8-flash-next/): Alibaba's Qwen3.8-Flash-Next is an open-weight Qwen4 architecture preview. Confirmed facts, rumored 125B-A6B specs, and what is still missing.
- [The first 3T-class open model is here. What Kimi K3 changes](https://indierouter.ai/blog/kimi-k3-open-frontier-model/): Moonshot shipped Kimi K3, a 2.8 trillion parameter open-weight model you can actually self-host. Here is what the largest open model yet means for India.
- [Your frontier model can be switched off by a foreign government](https://indierouter.ai/blog/sovereign-ai-access-dependency/): When the US ordered Anthropic to cut foreign access to its frontier models, AI sovereignty became real. Here is what Indian teams should do.
- [The real cost of LLM inference is the GPU you are not using](https://indierouter.ai/blog/llm-inference-cost-utilization/): Token prices have fallen roughly 600x since 2020, but real inference cost is set by GPU utilization, not the sticker price. Here is the math.
- [Qwen3.8-27B is the open-source model Indian teams can build on](https://indierouter.ai/blog/qwen-3-8-27b-open-source/): Qwen3.8 shipped twice under two licences: an Apache 2.0 27B that keeps vision, and a 2.4T flagship under a revenue-share licence. Which can you build on?
- [Qwen3.8 27B is live: fast vision-language AI in India](https://indierouter.ai/blog/qwen3-8-27b-launch/): Qwen3.8 27B is live on IndieRouter: 262K context, vision input, ₹40/M input and ₹200/M output, automatic prompt caching, served from India.
- [DeepSeek V4-Flash-Vision-Exp: multimodal without the vision surcharge](https://indierouter.ai/blog/deepseek-v4-flash-vision-exp/): DeepSeek's first vision model in the V4-Flash line matches V4-Flash on text and closes most of the gap to Opus-4.8, with no vision surcharge.
- [DPDP enforcement starts Nov 2026: what it means for AI teams in India](https://indierouter.ai/blog/dpdp-compliance-deadline-ai-teams/): India's data protection rules now have a phased enforcement timeline. Here's what the deadlines require and why your inference runs where it does.
- [IndieRouter is in open beta: DeepSeek V4 Flash is live on Indian GPUs](https://indierouter.ai/blog/indierouter-open-beta/): Every new account starts with ₹500 of free inference credit. Change one base URL and your OpenAI SDK runs on sovereign Indian infrastructure.

## Legal
- [Privacy](https://indierouter.ai/privacy/): DPDP-aligned data handling, zero-retention inference, and India-only processing.
- [Terms](https://indierouter.ai/terms/): Account, acceptable-use, billing, SLA, and liability terms for the IndieRouter API.
- [Refunds](https://indierouter.ai/refunds/): Refund and cancellation policy for prepaid credits, subscriptions, and UPI AutoPay mandates.