Seed1.6
ByteDance Seed / 豆包 · 2025-06-25 · 类别未确认
官方发布来源 · 按需求选型 · 查看覆盖缺口 · 归档与转录记录
Seed1.6
发布文将其定位为带自适应思考(AdaCoT)的多模态通用系列,称视觉任务追平甚至超过 Seed1.5-VL。已收录的 8 项数值均来自 Seed1.6 Base 列的知识与数理基础评测,亮点为 BBH 92.08 与 MMLU 88.83;多模态旗舰版本身未给出数值。
- 输入模态
- 文本 / 图像
- 上下文
- 256K
- 参数
- 230B-A23B 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: LLM evaluation table (rendered non-DOM component) · table: Seed1.6 Base table (groups: General Knowledge & Reasoning / STEM & Math & Coding) · row: MMLU · figure: images/05.jpg (native-resolution crop re-read 2026-09-01)
{
"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
}General Knowledge & Reasoning group. Table columns: Seed1.6 Base / LLaMA-4-Maverick Base / DeepSeek-V3 Base / Qwen3-235B-A22B Base / Seed1.5 Base. Page states external-model columns are sourced from the Qwen Technical Report (second-hand comparison). Same-table competitor cells (LLaMA-4-Maverick / DSV3 / Qwen3-235B / Seed1.5): 85.16/87.19/87.81/88.35. Vision read at native resolution.
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: LLM evaluation table (rendered non-DOM component) · table: Seed1.6 Base table (groups: General Knowledge & Reasoning / STEM & Math & Coding) · row: MMLU-Pro · figure: images/05.jpg (native-resolution crop re-read 2026-09-01)
{
"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
}General Knowledge & Reasoning group. Table columns: Seed1.6 Base / LLaMA-4-Maverick Base / DeepSeek-V3 Base / Qwen3-235B-A22B Base / Seed1.5 Base. Page states external-model columns are sourced from the Qwen Technical Report (second-hand comparison). Same-table competitor cells (LLaMA-4-Maverick / DSV3 / Qwen3-235B / Seed1.5): 63.91/59.84/68.18/66.47. Vision read at native resolution.
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: LLM evaluation table (rendered non-DOM component) · table: Seed1.6 Base table (groups: General Knowledge & Reasoning / STEM & Math & Coding) · row: SuperGPQA · figure: images/05.jpg (native-resolution crop re-read 2026-09-01)
{
"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
}General Knowledge & Reasoning group. Table columns: Seed1.6 Base / LLaMA-4-Maverick Base / DeepSeek-V3 Base / Qwen3-235B-A22B Base / Seed1.5 Base. Page states external-model columns are sourced from the Qwen Technical Report (second-hand comparison). Same-table competitor cells (LLaMA-4-Maverick / DSV3 / Qwen3-235B / Seed1.5): 40.85/41.53/44.06/36.81. Vision read at native resolution.
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: LLM evaluation table (rendered non-DOM component) · table: Seed1.6 Base table (groups: General Knowledge & Reasoning / STEM & Math & Coding) · row: BBH · figure: images/05.jpg (native-resolution crop re-read 2026-09-01)
{
"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
}General Knowledge & Reasoning group. Table columns: Seed1.6 Base / LLaMA-4-Maverick Base / DeepSeek-V3 Base / Qwen3-235B-A22B Base / Seed1.5 Base. Page states external-model columns are sourced from the Qwen Technical Report (second-hand comparison). Same-table competitor cells (LLaMA-4-Maverick / DSV3 / Qwen3-235B / Seed1.5): 83.62/86.22/88.87/88.36. Vision read at native resolution.
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: LLM evaluation table (rendered non-DOM component) · table: Seed1.6 Base table (groups: General Knowledge & Reasoning / STEM & Math & Coding) · row: GPQA-Diamond · figure: images/05.jpg (native-resolution crop re-read 2026-09-01)
{
"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
}STEM & Math & Coding group. Table columns: Seed1.6 Base / LLaMA-4-Maverick Base / DeepSeek-V3 Base / Qwen3-235B-A22B Base / Seed1.5 Base. Page states external-model columns are sourced from the Qwen Technical Report (second-hand comparison). Same-table competitor cells (LLaMA-4-Maverick / DSV3 / Qwen3-235B / Seed1.5): 43.94/41.92/47.47/45.25. Vision read at native resolution.
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: LLM evaluation table (rendered non-DOM component) · table: Seed1.6 Base table (groups: General Knowledge & Reasoning / STEM & Math & Coding) · row: GSM8k · figure: images/05.jpg (native-resolution crop re-read 2026-09-01)
{
"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
}STEM & Math & Coding group. Table columns: Seed1.6 Base / LLaMA-4-Maverick Base / DeepSeek-V3 Base / Qwen3-235B-A22B Base / Seed1.5 Base. Page states external-model columns are sourced from the Qwen Technical Report (second-hand comparison). Same-table competitor cells (LLaMA-4-Maverick / DSV3 / Qwen3-235B / Seed1.5): 87.72/87.57/94.39/89.99. Vision read at native resolution.
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: LLM evaluation table (rendered non-DOM component) · table: Seed1.6 Base table (groups: General Knowledge & Reasoning / STEM & Math & Coding) · row: MATH · figure: images/05.jpg (native-resolution crop re-read 2026-09-01)
{
"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
}STEM & Math & Coding group. Table columns: Seed1.6 Base / LLaMA-4-Maverick Base / DeepSeek-V3 Base / Qwen3-235B-A22B Base / Seed1.5 Base. Page states external-model columns are sourced from the Qwen Technical Report (second-hand comparison). Same-table competitor cells (LLaMA-4-Maverick / DSV3 / Qwen3-235B / Seed1.5): 63.32/62.62/71.84/66.18. Vision read at native resolution.
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: LLM evaluation table (rendered non-DOM component) · table: Seed1.6 Base table (groups: General Knowledge & Reasoning / STEM & Math & Coding) · row: MBPP · figure: images/05.jpg (native-resolution crop re-read 2026-09-01)
{
"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
}STEM & Math & Coding group. Table columns: Seed1.6 Base / LLaMA-4-Maverick Base / DeepSeek-V3 Base / Qwen3-235B-A22B Base / Seed1.5 Base. Page states external-model columns are sourced from the Qwen Technical Report (second-hand comparison). Same-table competitor cells (LLaMA-4-Maverick / DSV3 / Qwen3-235B / Seed1.5): 75.40/74.20/81.40/81.40. Vision read at native resolution.