GLM-4.6 · 官方评测披露
GLM-4.6 当前收录 10 项测试的官方记录。已完成列明材料的核查。
模型发布 · GLM
官方材料与核查范围
All official source bodies and directly linked evaluation material listed for this release were read in scope; no verifiable launch-day snapshot was obtained. 核对时间 。
当前资料不足以还原首发。 No verifiable first-release historical snapshot was obtained; current pages and repository content do not establish the full launch disclosure set.
- Z.AI New Released · 主文 · Official entry for GLM-4.6 states 2025-09-30.公布 未确认 · 采集 · 已核查 Official release-history body read for the named model/date; it is not a launch-day snapshot.
- GLM-4.6 | Z.AI Docs · 主文 · Current official GLM-4.6 page; benchmark chart and task labels checked; collected 2026-11-17.公布 未确认 · 采集 · 已核查 Saved official source body read in full; benchmark passages and directly linked evaluation material in scope checked.
- LLM Performance Evaluation: Agentic, Reasoning and Coding · 附件 · Official embedded chart downloaded as 本地留存副本 and visually read; exact score labels, 8 benchmark names and w/Tools annotations checked.公布 未确认 · 采集 · 已核查 Direct chart/image read visually; exact target labels transcribed without estimating from coordinates.
- Official GLM-4.6 CC-Bench V1.1 comparison chart · 附件 · Direct Z.AI documentation image; all four win/tie/loss rows visually read; local file 本地留存副本.公布 未确认 · 采集 · 已核查 Direct chart/image read visually; exact target labels transcribed without estimating from coordinates.
- GLM-4.6 official model card · 主文 · 本地2026-11-17官方卡片快照;2026-11-18完整目标列、内嵌图和原文脚注核对完成。公布 未确认 · 采集 · 已核查 卡片全文、目标模型完整正向数值列、适用脚注及图内标签核对完成;未从其他比较列借分。
配置
- GLM-4.6 official chart configuration · GLM-4.6; chart column is not labelled w/ Tools. Model serving, effort and precision are not disclosed. 未知项 Effort and serving endpoint are unknown.、Chart values have no printed units.
- GLM-4.6 chart result labelled w/ Tools · GLM-4.6; exact tool setup and harness are not described by the figure. 未知项 Tool protocol, effort, endpoint and precision are unknown.、Chart values have no printed units.
- GLM-4.6 · 档位未注明 · 官方模型卡目标列;逐行scaffold、工具、采样及预算条件见protocol。未由默认模式或推荐解码设置推断实际推理档位。 未知项 未逐项说明的指标单位、题集修订、服务精度和运行次数保持未知。
直接前代与后续发布
当前收录未确认直接前代。 官方关系依据
The selected official sources do not explicitly establish a direct predecessor; none is inferred.
测试记录 · 10
- AIME 25 · GLM-4.6模型卡;2025; full I/II subset not stated本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
- AIME 25 · GLM-4.6模型卡;2025; full I/II subset not stated; w/ Tools本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
- GPQA · Subset/version unspecified in GLM source本次发布有 2 条成绩记录,含 2 条精确值(可含历史记录)
- LiveCodeBench · v6; GLM source metric unit unspecified本次发布有 2 条成绩记录,含 2 条精确值(可含历史记录)
- Humanity's Last Exam · Version, tools and unit unspecified in chart本次发布有 2 条成绩记录,含 2 条精确值(可含历史记录)
- BrowseComp · Protocol/version unspecified in GLM source本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
- SWE-bench Verified · GLM-4.6模型卡;Verified本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
- Terminal-Bench · 2.0; GLM source does not name score/mean-reward metric本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
- τ²-Bench · Weighted aggregate; release/version unspecified本次发布有 1 条成绩记录,含 1 条精确值(可含历史记录)
- CC-Bench · V1.1; 74 real-world coding tests in Claude Code; exact task release unspecified本次发布有 12 条成绩记录,含 12 条精确值(可含历史记录)