LLM Leaderboard Top 10: Which Open Models Can You Actually Run Locally? (Nov 2026)
How to read the LLM leaderboard top 10: a 2026-11 snapshot of LMArena’s open-license board plus Artificial Analysis, with Q4 size estimates for local use.
-
-
Tencent Hy4 Fact Check, First 48 Hours: What Can Be Verified and What Is Only Self-Reported
Tencent Hy4 launched 2026-11-04 with 770B parameters and weights promised soon. Which claims can be checked, which are self-reported, and the memory it needs.
-
大模型排行榜前十名怎么看：开源榜 Top 10 与你的显卡能跑哪一个（2026-11 快照）
大模型排行榜前十名怎么看？整理 LMArena 开源榜与 Artificial Analysis 的快照，估算 Q4 体积，告诉你前十名里哪些能在本地跑。
-
腾讯混元 Hy4 发布 48 小时事实核查：哪些说法能核对，哪些只是自报
腾讯混元 Hy4 于 2026-11-04 发布：770B 参数、权重“近期”开源。逐条核对哪些说法可查、哪些是自报，并估算本地运行所需内存。
-
How to Fact-Check Leaked Model Claims and “Cracked” Releases: A 7-Step Method (numbat and GSQ-RCO Examples)
Leaked model claims are hard to judge. A 7-step method: sources, metadata, arithmetic, contradictions, definitions, replication and safe handling, two cases.
-
Refusal Rate Benchmarks Explained: Over-Refusal vs Harmful Compliance (XSTest, OR-Bench, SORRY-Bench, GB/T 45654)
Refusal rate benchmarks explained: the two numbers behind a refusal rate, XSTest, OR-Bench and SORRY-Bench sizes, GB/T 45654’s 95%/5% rule and intervals.
-
Uncensored Local LLM: Legal and Compliance Risks for Individuals, Distributors and Companies
Is an uncensored local LLM legal? Compare China and EU rules, licence duties and a self-check list for personal use, public distribution and company deployment.
-
大模型拒答率怎么测、怎么看：过度拒答与有害放行（XSTest、OR-Bench、SORRY-Bench 与 GB/T 45654）
拒答率怎么测？讲清过度拒答与有害放行两个数字、XSTest、OR-Bench、SORRY-Bench 等测试集规模、GB/T 45654 的 95%/5% 要求与置信区间算例。
-
本地大模型解除限制合法吗？无审查模型的合规与法律风险：个人、分发者与企业的边界
本地大模型解除限制合法吗？梳理无审查模型在个人使用、公开分发、企业部署三种角色下的中国与欧盟规则、许可证义务与合规自查清单。
-
模型泄露核查 7 步法：怎么判断“权重泄露”“破解版”消息是真是假（以 numbat、GSQ-RCO 为例）
模型泄露核查 7 步法：从来源、元数据、算术自洽、自述矛盾、口径、独立复现到安全处置，用 numbat 与 GSQ-RCO 两个案例演示如何辨别真假。
