网友质疑 Artificial Analysis 榜单偏袒美国模型,征求替代评测站
thiswebthisweb · reddit · 2026-10-28
A Reddit user complains that Artificial Analysis no longer lists GLM-5.3 Flash — which they call one of the best and cheapest-per-intelligence models — and has also hidden Qwen 3.8 Flash and other cost-effective models, suspecting ranking changes that favor US models. They ask for alternative model benchmarking/comparison sites.
Beyond the request itself, the post highlights a real concern: changes in how third-party leaderboards curate and rank models can significantly affect the visibility of open-weight and Chinese models.
所属事件:AA 智能指数升级 v5.0,中国开源模型集体掉分引中立性质疑(7 条相关)→
「模型」频道最新
- 人设提示词让模型在 SIMNER 基准一律锁 35%,换模型也没用 — amplifiedamp · 2026-10-29
- OpenAI Terra 6.1 评测：便宜模型配对调用，跑赢基准测试看不见的经济账 — IamNotGorbachev · 2026-10-29
- qwen-code 合并 PR：Qwen4 全系上下文窗口定为 100 万 tokens — haider1 · 2026-10-29
- 用户质疑 Claude 周额度未用先扣：重置后 2% 莫名消失 — umeshonai · 2026-10-29
- 「人格神经元」遭质疑:改几个浮点数就能击穿对齐人格 — aiamblichus · 2026-10-29
- 社区为 Qwen4 提愿望清单:512K 上下文与「说人话」呼声最高 — perelmanych · 2026-10-29