Topic
Open-weight models
AI models whose weights are released publicly for anyone to download, run and fine-tune.
Nvidia buys the open-model commons: inside the $12.93bn Hugging Face deal
The company that sells the shovels now owns the map. What changes when the dominant accelerator vendor controls where open models are found, compared and deployed.
DeepSeek's V4-Pro went GA in silence, then got a lot more expensive
A quiet pricing-page edit and a four-day-later price rise of up to 1,100% reveal more about DeepSeek's cost position than either fact does alone.
Kimi K3 and the open-weights escalation: what 2.8 trillion parameters actually buys
Moonshot published the largest set of open weights ever released. The benchmark story is genuinely strong; the deployment story is why the frontier labs can still sleep.
Meituan's LongCat-2.0: the first trillion-parameter model built without Nvidia
A 1.6-trillion-parameter open model is the smaller story. The compute stack underneath it — 50,000 domestic chips, no Nvidia — is the one that changes the map.
The end of tokenmaxxing: why AI buyers are routing to cheaper open-weight models
The assumption that intelligence could be metered by the token, and that usage would only rise, has quietly become negotiable.
The 2026 open-weight surge, explained
A year of downloadable models that narrowed the gap with the closed frontier — mostly from China.
Qwen and the rise of open-source AI from China
One of the most-downloaded model families anywhere — and the strategy behind it.
Kimi K2.7-Code and the metronome of Chinese open weights
Another month, another capable open model from a Chinese lab. The cadence is the story.
Mistral's €20bn raise and Europe's sovereign-AI bet
Nearly doubling its valuation in months — the thesis is as much geopolitical as financial.
MiniMax M3: the open-weight frontier, with an asterisk on 'open'
A genuinely ambitious model — frontier coding, 1M context, multimodal — shipped before its weights did. Here is what that means.
DeepSeek V4: what the architecture shift actually signals
A 1.6-trillion-parameter open model at a fraction of the closed-frontier price — but the engineering story beneath it is more durable than the benchmark.
Voxtral: Mistral's open-weight voice model is a structural bet against the incumbents
A 4B TTS that runs on one consumer GPU — and rewrites the value proposition for the whole voice AI market.