Hunyuan-A13B-Instruct / Hunyuan-A13B-Pretrain / Hunyuan-A13B-Instruct-FP8 / Hunyuan-A13B-Instruct-GPTQ-Int4
Tencent / 腾讯混元 · 2025-06-27 · 类别未确认
Hunyuan-A13B-Instruct
腾讯混元以「细粒度 MoE、80B 总参/13B 激活」开源 A13B 系列,Instruct 为指令版(256K 上下文,/think 与 /no_think 双模式,默认慢思考)。已收录 21 项 Instruct 表评测覆盖知识、数学、代码与 Agent 工具调用:亮点 AIME24 87.3、GPQA Diamond 71.2。
- 输入模态
- 文本
- 上下文
- 256K
- 参数
- 80B-A13B MoE
- 价格(每百万 tokens)
- 无此口径报价;不等于免费
本变体的评测证据
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: aime24
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: aime-25
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: math
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: gpqa
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: olympiadbench
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: olympiadbench not yet in data/benchmarks/.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: lcb
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: fullstackbench
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: fullstackbench not yet in data/benchmarks/.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: artifactsbench
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: bbh
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: drop
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: zebralogic
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: ifeval
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: sysbench
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: sysbench not yet in data/benchmarks/.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: lengthctrl
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: lengthctrl not yet in data/benchmarks/.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: insctrl
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: insctrl not yet in data/benchmarks/.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: complexnlu
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: complexnlu not yet in data/benchmarks/.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: word-task
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: word-task not yet in data/benchmarks/.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: bfcl
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: tau-bench
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: complexfuncbench
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: complexfuncbench not yet in data/benchmarks/.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (Hunyuan-A13B-Instruct, topic-grouped table) · table: Instruct benchmark table (Topic / Bench) · row: c3-bench
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: c3-bench not yet in data/benchmarks/.
Hunyuan-A13B-Pretrain
Hunyuan-A13B-Pretrain 为该开源系列的基座模型(Base 表以 TRT-LLM 后端评测)。已收录 14 项 Base 表评测:亮点 MMLU 88.17、GSM8K 91.83。
- 输入模态
- 文本
- 上下文
- 官方资料未说明
- 参数
- 80B-A13B MoE
- 价格(每百万 tokens)
- 无此口径报价;不等于免费
本变体的评测证据
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: mmlu
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: mmlu-pro
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: mmlu-redux
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: bbh
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: supergpqa
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: evalplus
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: multipl-e
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: mbpp
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: crux-i
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: crux-i not yet in data/benchmarks/. Page prints the row as CRUX-I.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: crux-o
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: crux-o not yet in data/benchmarks/. Page prints the row as CRUX-O.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: math
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: cmath
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: cmath not yet in data/benchmarks/. Page prints the row as cmath.
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: gsm8k
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Benchmark (base models) · table: base-model benchmark table · row: gpqa
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}Hunyuan-A13B-Instruct-FP8
Hunyuan-A13B-Instruct-FP8 为 Instruct 的 FP8 静态量化版(AngelSlim)。已收录 4 项评测:AIME24 86.7、GSM8K 94.01、BBH 88.34、DROP 91.1。
- 输入模态
- 文本
- 上下文
- 官方资料未说明
- 参数
- 80B-A13B MoE
- 价格(每百万 tokens)
- 无此口径报价;不等于免费
本变体的评测证据
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: FP8 Quantization / FP8 Benchmark · table: FP8 benchmark table · row: aime24
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: FP8 Quantization / FP8 Benchmark · table: FP8 benchmark table · row: gsm8k
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: FP8 Quantization / FP8 Benchmark · table: FP8 benchmark table · row: bbh
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: FP8 Quantization / FP8 Benchmark · table: FP8 benchmark table · row: drop
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}Hunyuan-A13B-Instruct-GPTQ-Int4
Hunyuan-A13B-Instruct-GPTQ-Int4 为 Instruct 的 GPTQ W4A16 量化版(AngelSlim)。已收录 5 项评测:亮点 AIME24 86.7、OlympiadBench 84.0。
- 输入模态
- 文本
- 上下文
- 官方资料未说明
- 参数
- 80B-A13B MoE
- 价格(每百万 tokens)
- 无此口径报价;不等于免费
本变体的评测证据
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Int4 Quantization / Int4 Benchmark · table: Int4 benchmark table · row: olympiadbench
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}new-benchmark: olympiadbench not yet in data/benchmarks/ (also referenced by the Instruct table row).
归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Int4 Quantization / Int4 Benchmark · table: Int4 benchmark table · row: aime24
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Int4 Quantization / Int4 Benchmark · table: Int4 benchmark table · row: gsm8k
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Int4 Quantization / Int4 Benchmark · table: Int4 benchmark table · row: bbh
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}归一化读数:不计算;样本及方差不完整,无法计算置信区间。
原文位置与完整协议
heading: Int4 Quantization / Int4 Benchmark · table: Int4 benchmark table · row: drop
{
"harness": null,
"tools": null,
"shots": null,
"reasoning_effort": null,
"temperature": null,
"top_p": null,
"token_budget": null,
"turn_limit": null,
"time_limit": null,
"run_count": null,
"aggregation": null,
"judge": null
}