#DeepSeek

The hard part of chip independence was never just making the chip but getting developers to write code for it, and DeepSeek is solving that toughest software problem for Huawei.

Open models are free; the fight is over where developers first download them, because that's often where they first pay for compute.

Chinese model makers are moving from chasing a benchmark score to selling one model at several price points, making pricing design as important as capability.

Competition among agent tools is shifting from whose model codes better to whose installer is easier and whose plugins are easier to find; the desktop is the new battleground.

A model business must show both that customers will keep paying and that each extra dollar of revenue does not require an equivalent increase in compute spending.

The bottleneck in agent training is shifting from GPUs to CPUs and environment scheduling, and by publishing DSec, DeepSeek signals it no longer sees this layer as a moat.

The lab famous for saving compute is now quoting trillions of parameters; Chinese model competition has switched its unit from efficiency to scale.

Rejecting restraint at the podium while inviting companies to discuss risk in the council — global AI governance received two opposite signals in one week.

$3 million for an open model tied with Grok 4.7 — a phone maker is repricing the entry fee to frontier capability.

The test of domestic compute moving from inference to frontier training is sustained cluster performance, not a shipment announcement.

Token share shows where models are being tried and called; commercial share still depends on who pays and at what price.

AI may first consume the repetitive manual optimization time of top engineers, rather than the engineering job itself.

The confirmed desktop progress concerns build and update infrastructure; general installer availability and comparable task performance require separate verification.

Version numbers no longer mark a capability jump; unit cost is the column that has to change.

A company valued at RMB 500 billion filling its finance seat usually means capital moves are on a schedule.

New-model support has become an hourly contest; a usable day-one image can influence developer choice as much as hardware specifications.

Four new voices are a confirmed product change; there is no evidence yet that the underlying model is new, so the feature and a model release must remain separate.

DeepSeek is debuting its new architecture in a low-priced Flash model, leaving real-world task costs to determine whether the efficiency gain holds.

DeepSeek is pushing model upgrades into the agent runtime at the same time, giving developers a broader product interface alongside the new model.

IPO preparations provide a potential financing route. Formal revenue, cost and capital-demand disclosures would reveal more about the economics of the model business.

Backend substitution reduces migration work but leaves developers needing version identification and regression tests. Lower bills and stable task performance need to be accepted together.

The limited beta offers a window to test the new architecture; improvements in capability, speed and cost each need measurement on the same real tasks.

Caching receives the largest cut, but actual savings depend jointly on the cache-hit rate, output length and share of work performed at peak times.

User growth puts backend engineering in the foreground; the payoff from hiring should appear in more reliable task delivery, not merely a larger team.

Training on Nvidia, inference on domestic silicon — where that line sits is the real progress bar for China's compute self-sufficiency.

160,000 Ascend chips under one lab's name moves domestic compute from "usable" to "bet on by a leading lab."

Revenue up fourfold, stock down sixty percent — that gap is the market repricing the story told by China's model companies.

305B weights under MIT put the first squeeze not on OpenAI but on every domestic multimodal API billing by the call.

Once the racks are self-built, DeepSeek's cost advantage stops being an algorithm question and becomes a capex question.

The watershed this time isn't benchmark scores — it's that a frontier-scale model has, for the first time, proven it can survive without NVIDIA's supply rhythm.

The one-ninth training-cost figure says more than any benchmark score about which battle the next-generation Qwen aims to win.

The revenue-share talks were never just about the money — they're about whether Chinese models get a named, legitimate supplier identity on US clouds.

The 82.9% API gross margin proves the inference business itself can make money; the question has always been who pays for the free half.

A chipmaker building the strongest open-source model itself was never about selling the model — it's about becoming the default option that routes everyone's inference onto its own hardware.

Where the pricing sheet loosens is usually where overcapacity first surfaces.

A three-month limited-time price cut is not a promotion — it's OpenAI leaving itself an opening to adjust prices again at any time.

Comparison numbers produced on a self-run Harness are only worth as much as how quickly third parties can reproduce them.

With self-developed and open-source models sitting side by side in the internal test interface, Tencent has no intention of selecting just one of the two paths.

Converting images to text and feeding the result to text-only models paves the way with engineering before model capabilities catch up.

When APIs start billing by peak and off-peak windows, compute officially becomes a utility.

The real objective of open-sourcing a runtime is getting others' engineering habits to take root in your abstraction.

Beyond the model, the agent runtime is becoming the next layer to be conquered by open source.

List price cut to a tenth, cache prices up tenfold — DeepSeek has turned a price war into a price-structure war and shifted the full burden of price comparison onto developers.

With models free and frameworks open-sourced, the truly scarce assets left in the agent race are usage entry points and dominance of the plugin ecosystem.

Intelligence gap of 1 point, price gap of 8x — the high-priced tier needs a new justification.

The four-times-cheaper flagship model carries its own price-increase warning; cost models need to leave a line for it.

Usage at 6%, revenue at 11.4% — premium models sell task value, not call counts.

Between shipping a model and shipping a product sits an organization that can interview every day and open an official account.

Redoing post-training once was enough to overtake the premium version, showing the bar for agent capability is shifting from "stacking parameters" to "refining the recipe" — precisely the stage where the open-source camp is closing the gap fastest.

DeepSeek's stake in Unitree turns the convergence of "brain" and "body" from a forum topic into an equity structure.

The easiest way for a hardware company to get a brain is to make the brain-maker a shareholder.

The price slasher is raising prices itself — the ledger behind rock-bottom pricing no longer balances.

Eight domestic chips adapted on the same day says more about what MiniMax has a grip on than a No. 1 spot on the leaderboard.

A model company is paying for robot-body equity — upstream and downstream in embodied AI are crossing into each other's turf.

The coding-model price war is on — the first shot lands on competitors' margins, and the ammunition is data.

RMB 1 trillion of intent chases an RMB 50 billion quota — money was never the scarce asset.

What counts is no longer any one release, but the once-a-month release calendar.

When benchmark parity costs a tenth of the price, the middle-tier API has no story left to tell.

Per-million-token pricing is losing its reference value; what counts is the total bill for getting a task done.

The moment a cheap model catches up to a premium one, what gets eliminated isn't the premium model — it's model routing as a business.

Open source isn't a margin concession — it's moving the distribution channel from someone else's console into your own hands.

Run the same model twice, and the second pass is worth 10 extra points — post-training is becoming a more valuable asset than parameter count.

Power can be bought; cards may not be — two different ledgers.

One side is handing out weights for free, the other is more than doubling subscription prices — both companies are betting on the same thing: the money isn't in the model.