Open weights stopped chasing a champion and built a toolbox

The single-best-open-model question has quietly lost its meaning. The open field now has an all-rounder and a set of specialists — which is what a mature market looks like, not a race.

GLM-5.2 has emerged as the strongest all-round open-weight model — MIT-licensed, long context — with Kimi K2.7 Code, Gemma 4 12B, and Nemotron 3 Super as clear specialists. An all-rounder plus fit-for-purpose specialists is the signature of a matured ecosystem, not a scramble of single benchmarks.

Choosing an open model became an engineering decision

Because the leaders now diverge by task. Qwen leads reasoning and coding, DeepSeek leads deep math, Llama 4 Scout leads long context at 10M tokens, Gemma leads local. You no longer pick a champion; you match a tool to a job. That most real workloads have a specific shape is exactly why a toolbox beats a single hero model.

The pressure this puts on the closed premium

When the best open all-rounder is MIT-licensed and specialists cover every major use case at a fraction of the cost, the closed labs must justify their price workload by workload — which is the same pressure driving the frontier price war. Open weights are no longer the fallback. They're the default, and the licences make them deployable.

The market's shape tells the story: not one open model rising, but a whole capable, permissively-licensed toolkit settling into place beneath the closed frontier.

Techsy — Best open-source LLMs: July 2026 leaderboard → · Fireworks — Best open source LLMs in 2026 →