智谱发了 GLM 5.2，Benchmark 接近 Opus，成本才一毛八，还有更便宜的 M3 和升级的 Kimi 2.7，做浏览器智能体有得选了。
GLM 5.2 在 BrowserCode 基准上达到接近 Opus 级别分数，单次任务成本仅 0.18 美元。Minimax M3 取得 Sonnet 级别分数，输入价格 0.30 美元，比 DeepSeek V4 Pro 更便宜。Kimi k2.7 相比 k2.6 提升 9%，但被 M3 超越。这些开源权重模型首次在浏览器智能体任务中追平甚至超过 GPT 5.5 等闭源模型。
Open-weights models have officially caught up We tried GLM 5.2 in BrowserCode > Near Opus-level ...
Open-weights models have officially caught up We tried GLM 5.2 in BrowserCode > Near Opus-level score > Cheapest model to date This task took $0.18 Try it now ↓ Your browser does not support the video tag. 🔗 View on Twitter Alexander Yue @Alezander907 GLM 5.2 is a huge improvement for browser agents, offering near opus level score, beating GPT 5.5 Minimax M3 is a sonnet level score at just $0.30 input, my new best value model (cheaper than deepseek v4 pro) Kimi k2.7 is a +9% improvement from k2.6 but is outclassed by M3 🔗 View Quoted Tweet 💬 2 🔄 0 ❤️ 22 👀 1388 📊 6 ⚡