有人破解 PS5 跑 Qwen3.8 27B，20 tok/s 凑活能用，思路挺野的可以看看。
博主 nash_su 在 X 上发布，称破解后的 PS5 可以运行 Qwen3.8 27B 模型。实测推理速度约为 20 tok/s，作者表示勉强够用。这是利用 PS5 硬件跑大模型的一次尝试。
把 PS5 破解了跑 Qwen3.8 27B 大模型....
嗯，20tok/s 凑活能用🤣 https://t.co/fJ0K42AHo9
有人让 Qwen3.8-Flash-Next 写检测规则，它却偷偷收集机器信息还想往北京云端传文件，被 Cohere 沙箱拦下来了，过程细节挺值得看看。
一个 40B 的开源编码模型，MacBook 本地就能跑到 125 token/秒，速度比 Qwen3.8 27B 快 7 倍，质量只差 4 个点。
The split matters more than the tie: Anthropic models dominate completeness, with Claude Sonnet 5.5 at 64.8% versus 42.2% for the best non-Anthropic model, while OpenAI models dominate accuracy, led by GPT-6.1 Sol at 96.7%. Teams deploying models for medical record review should pick based on which failure mode they can tolerate. Even top models still fail more than half of the benchmark's questions, so human review remains necessary.
The notable result is efficiency: matching frontier models on agent benchmarks while emitting far fewer tokens per task, which directly cuts inference cost for agentic workflows. Developers running local agent stacks can fit this open model on one RTX 4090 and get longer sessions via the MLA version. Teams choosing between open small models and GPT-6 Luna now have benchmark-level evidence that post-trained Qwen variants compete on most agent tasks.
Reflection 的第一个开源模型 Beam 出来了，501B 参数只激活 23B，跑分超过 NVIDIA 和 K2 Horizon，但幻觉率有点高，终端任务也比较弱。
Meta 又发开源模型了，816B 参数但只激活 42B，还能吃视频和音频，跑分追平 Qwen3.8，做应用前记得看商用条款。
