Qwen3.8-Max · Terminal-Bench(2.1; evaluated with Claude Code at avg@10, 5-hour timeout, max_tokens=131,072; metric scale unspecified) 86.6 · 附件证据(ev-qwen-3-8-max-api-01)

Qwen3.8-Max 的 Terminal-Bench · 2.1; evaluated with Claude Code at avg@10, 5-hour timeout, max_tokens=131,072; metric scale unspecified 官方出处。QwenCloud API: qwen3.8-max · 86.6。见《Qwen3.8-2.4T-A95B official model card with explicit Qwen3.8-Max comparison column》Benchmark Results table / Coding Agent / Terminal Bench 2.1 / Qwen3.8-Max column。证据 ev-qwen-3-8-max-api-01。

原文与定位

Qwen3.8-Max · Terminal-Bench(2.1; evaluated with Claude Code at avg@10, 5-hour timeout, max_tokens=131,072; metric scale unspecified),QwenCloud API: qwen3.8-max配置,Reported score 86.6(具体成绩公布时间未确认)。见《Qwen3.8-2.4T-A95B official model card with explicit Qwen3.8-Max comparison column》Benchmark Results table / Coding Agent / Terminal Bench 2.1 / Qwen3.8-Max column。 查看成绩条件

| | Opus 4.8 | Fable 5 | GPT 5.6 Sol (max) | Qwen3.7-Max | Qwen3.8-Max | Coding Agent | Terminal Bench 2.1 | 84.6 | 84.6 | 88.8 | 74.5 | 86.6 | 2. Terminal Bench 2.1: Evaluated with Claude Code (avg@10), using a 5-hour timeout and max_tokens=131,072. For all other models, we report the best published score across harnesses: Claude Opus 4.8 and Claude Fable 5 with Terminus 2 from Artificial Analysis (https://artificialanalysis.ai/evaluations/terminalbench-v2-1); GPT-5.6 Sol with Codex (https://openai.com/index/previewing-gpt-5-6-sol/).
官方材料
Qwen3.8-2.4T-A95B official model card with explicit Qwen3.8-Max comparison column
定位
附件 · Benchmark Results table / Coding Agent / Terminal Bench 2.1 / Qwen3.8-Max column
材料版本
2026-11-17当前材料快照;完整读取Qwen3.8-2.4T-A95B官方模型卡正文和比较表的全部32行、相关脚注;仅录明确Qwen3.8-Max列。除脚注明示BT/Elo的两项外,表中分数未标单位。
版本公布
未确认
披露时间
未确认 · 后补披露
采集时间
历史依据
暂无可靠历史版本依据
引用文本

Qwen3.8-Max · Terminal-Bench(2.1; evaluated with Claude Code at avg@10, 5-hour timeout, max_tokens=131,072; metric scale unspecified),QwenCloud API: qwen3.8-max配置,Reported score 86.6(具体成绩公布时间未确认)。官方出处 Qwen3.8-2.4T-A95B official model card with explicit Qwen3.8-Max comparison column(附件,2026-11-17当前材料快照;完整读取Qwen3.8-2.4T-A95B官方模型卡正文和比较表的全部32行、相关脚注;仅录明确Qwen3.8-Max列。除脚注明示BT/Elo的两项外,表中分数未标单位。)Benchmark Results table / Coding Agent / Terminal Bench 2.1 / Qwen3.8-Max column,https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B。Benchlog https://benchlog.lekshome.cc/evidence/qwen-3-8-max-api/ev-qwen-3-8-max-api-01/

关联成绩与条件

Qwen3.8-Max · Terminal-Bench 成绩与配置
指标与成绩配置与评测条件证据关系
Reported score86.6原文 86.6 · 单位 未标注
公布 未确认
QwenCloud API: qwen3.8-max
设置
Hosted QwenCloud API; official model page code uses model="qwen3.8-max". The overview describes Qwen3.8-Max as based on Qwen3.8-2.4T-A95B with vision input, non-thinking support, 1M-token context by default, and built-in tools; this is a distinct API product configuration, not the open-weight checkpoint.
未知项
The table does not identify an exact serving revision, quantization, inference stack, or per-row decode settings.、Qwen3.8-Max supports non-thinking; the table does not state whether thinking was enabled for these results.、For rows without a footnote, the benchmark protocol and metric scale are not stated; unknown units and direction remain unknown.
工具
工具条件未记录
推理预算
输出上限 131072 tokens
框架
Claude Code
协议
Claude Code; avg@10; 5-hour timeout; max_tokens=131,072
条件
Current model-card comparison table collected 2026-11-17; exact first disclosure date is not stated. Only the Qwen3.8-Max column is recorded.

直接支持该成绩

在排名中查看此成绩