Notes on sovereign inference

Ling-3.1-flash: a 560B open model built mostly on linear attention

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.

Series · Start here

Eight H200s: the inference campaign, in 15 parts

Follow our H200 NVL campaign from the first topology check to a million-token conversation: fifteen engineering stories, measured results and failures.

DeepSeek V4.1 Flash is live on IndieRouter, and cheaper than V4

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

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

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

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

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

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

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

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

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

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

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

Every new account starts with ₹500 of free inference credit. Change one base URL and your OpenAI SDK runs on sovereign Indian infrastructure.