Open-weights LLMs are either 5 months behind forever, or caught up by Christmas — pick your benchmark
One chart says the gap to closed models hits zero on December 3, 2026. Another, across 18 benchmarks, says the lag has sat flat at five months the whole time. The article's real subject is how slippery “how good is it” has become — and the thread says the framing misses the point.
On a single intelligence index the open-vs-closed gap projects to zero by Christmas; spread across 18 benchmarks it's stayed near five months for two years. Coding is the one clear mover, from 15 months behind to one or two.
The thread's reframe: the number doesn't matter, because closed access can be revoked and an open weight can't be put back in the bag — whatever capability it reaches stays reachable forever.
The same data, two opposite headlines
The doubleword.ai piece makes a careful, almost self-defeating point: the gap between open-weights and closed-source LLMs depends entirely on which yardstick you hold up. One intelligence index, extrapolated, projects the gap reaching zero around December 3, 2026 — “the open source singularity by Christmas.” But a boxplot across 18 benchmarks shows the average lag “almost completely flat, at just under 5 months for the entire period.” The lone exception is coding, which “has gone from 15 months behind to only a month or two behind.” The article's honest conclusion is that the gap is neither clearly closing nor widening — the exercise mostly exposes “the difficulty of measuring LLM quality.” Even the article's visualizations drew complaints; more than one reader found them “too noisy and dense … lines all over the place.”
“You can't put an open weight back in the bag”
The thread's sharpest move was to declare the whole metric beside the point. One commenter reframed it around permanence, not parity:
“Access to Fable can be removed. I don't see how an open weight model can ever be put back into the bag though.”
Given the week's news (frontier models gated to approved companies), that argument carried weight. Whatever an open model reaches, another reader added, “those [capabilities] stay forever … That can't be said for API-based models where a provider can sunset models whenever they feel like.”
The dependency the optimists keep underplaying
The counter-anxiety was about the spigot. One reader warned that open weights are mostly “the result of philanthropy by some private org … The spigot can be turned off at any time,” and another suspected it already has at Alibaba's Qwen for the smaller, home-runnable sizes. And before any of that, an ex-Meduza CTO flagged the article's own sloppiness: it “confuses open source models with open weights models — not the same thing,” a distinction the title quietly erases. The fight over whether the gap is five months or zero, in the end, matters less than who owns the faucet.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
The essay arguing that if AI does everything, even the billionaires become pets
Fernando Borretti follows one chain of logic to a bleak end: when machines do all cognitive and physical work, workers, the wealthy, and even heads of state all become economically unnecessary — and unnecessary people don't keep their power. “Sam Altman? How many divisions does he have?”