跳至正文

从目标出发,找到产品、方案与交付方法。

语言
全球站 / Global描述目标
超长上下文大海捞针400 tasks (% Multi-Needle)

MRCR 256K+ 多针检索基准

在 128K 至 2M 上下文中注入多组矛盾与时序更新事实,检验模型长程检索抗衰减能力。

参考基线快照(非实时遥测)· GEO 摘要CN · 2026-11-21

MRCR 256K+ 多针检索基准(CN 参考基线)前三:deepseek/deepseek-v4-1-flash / minimax/minimax-m3-1-flash / zhipu/glm-5-3-flash;峰值标价性价比 openai/gpt-6-luna(507.8/$)。分叉预警:openai/gpt-6-2-astra 属高基准参考延迟退化,openai/gpt-5-mini 属高性价比黑马(非实时遥测)。

口径说明:排行、标价成本、参考延迟与基准分来自静态公开规格与参考基线(非实时网关流量,未经第三方 ZDR 审计签章的节点保持 Unverified)。

评测任务规模400
指标口径% Multi-Needle
评测方法学

在 64K / 256K / 1M 分档窗口中评估精确检索率及前缀缓存加速后的延迟稳定性。

MRCR 256K+ Long Context 适配模型排行 (28)

28 个模型
#模型与实验室流形评分混合标价 / 1M参考速率与基准上下文 / 输出窗首选节点分叉与 ZDR 审计
1
@deepseek·deepseek/deepseek-v4-1-flash
90.3
100% conf
$0.24
392.1 lift/$
168 tok/s
ref 145ms TTFT · 94.1 bench
1M
max out 384K
deepseek-official
共识领跑Unverified
2
@minimax·minimax/minimax-m3-1-flash
90.1
100% conf
$0.24
381.7 lift/$
172 tok/s
ref 138ms TTFT · 91.6 bench
1M
max out 131K
alibaba-bailian
共识领跑Unverified
3
@zhipu·zhipu/glm-5-3-flash
89.9
100% conf
$0.22
410.0 lift/$
162 tok/s
ref 140ms TTFT · 90.2 bench
1M
max out 66K
zhipu-bigmodel
共识领跑Unverified
485.3
100% conf
$0.48
195.4 lift/$
124 tok/s
ref 215ms TTFT · 93.8 bench
1M
max out 131K
alibaba-bailian
共识领跑Unverified
5
@deepseek·deepseek/deepseek-v4-pro
80.4
100% conf
$0.92
105.6 lift/$
96 tok/s
ref 285ms TTFT · 97.6 bench
1M
max out 384K
deepseek-official
共识领跑Unverified
6
@minimax·minimax/minimax-text-01-preview
79.8
92% conf
$0.46
183.7 lift/$
72 tok/s
ref 390ms TTFT · 84.5 bench
1M
max out 16K
alibaba-bailian
稳健基线Unverified
7
@qwen·qwen/qwen3-coder-480b
79.4
100% conf
$0.64
147.2 lift/$
98 tok/s
ref 235ms TTFT · 94.2 bench
256K
max out 33K
alibaba-bailian
共识领跑Unverified
877.5
100% conf
$1.59
59.6 lift/$
96 tok/s
ref 255ms TTFT · 94.6 bench
1.0M
max out 66K
volcengine-ark
稳健基线Unverified
9
@zhipu·zhipu/glm-4-plus
77.1
98% conf
$0.62
143.8 lift/$
94 tok/s
ref 240ms TTFT · 88.6 bench
128K
max out 16K
zhipu-bigmodel
稳健基线Unverified
1076.9
100% conf
$2.10
46.4 lift/$
102 tok/s
ref 260ms TTFT · 97.4 bench
1M
max out 131K
alibaba-bailian
稳健基线Unverified
1176.9
100% conf
$2.00
47.9 lift/$
98 tok/s
ref 245ms TTFT · 95.8 bench
1M
max out 131K
zhipu-bigmodel
稳健基线Unverified
12
@doubao·doubao/doubao-1-5-pro-256k
75.7
77% conf
$0.47
189.2 lift/$
86 tok/s
ref 280ms TTFT · 88.9 bench
256K
max out 16K
volcengine-ark
稳健基线Warn
13
@qwen·qwen/qwen3-8-max
75.5
100% conf
$2.80
34.1 lift/$
94 tok/s
ref 270ms TTFT · 95.6 bench
1M
max out 66K
alibaba-bailian
稳健基线Unverified
14
@moonshot·moonshot/kimi-k3-1-pro
74.6
100% conf
$3.60
26.8 lift/$
92 tok/s
ref 310ms TTFT · 96.6 bench
1M
max out 131K
alibaba-bailian
稳健基线Unverified
15
@moonshot·moonshot/kimi-k2-thinking
74.6
99% conf
$0.96
97.0 lift/$
78 tok/s
ref 360ms TTFT · 93.1 bench
256K
max out 33K
alibaba-bailian
稳健基线Unverified
1660.9
100% conf
$0.18
507.8 lift/$
192 tok/s
ref 132ms TTFT · 91.4 bench
1.1M
max out 64K
openai-direct
共识领跑Unverified
17
@google·google/gemini-3-flash
59.9
100% conf
$0.24
380.0 lift/$
185 tok/s
ref 150ms TTFT · 91.2 bench
1M
max out 66K
google-vertex
共识领跑Unverified
1854.8
100% conf
$0.40
226.0 lift/$
148 tok/s
ref 170ms TTFT · 90.4 bench
400K
max out 64K
openai-direct
高性价比黑马Unverified
19
@google·google/gemini-2-5-flash-lite-exp
51.9
91% conf
$0.51
165.0 lift/$
108 tok/s
ref 250ms TTFT · 84.5 bench
1M
max out 33K
google-vertex
共识领跑Unverified
20
@meta-llama·meta-llama/llama-4-scout-17b
51.8
80% conf
$0.31
261.9 lift/$
96 tok/s
ref 295ms TTFT · 81.2 bench
1M
max out 16K
together-turbo
稳健基线Warn
2151.3
100% conf
$0.98
97.2 lift/$
118 tok/s
ref 225ms TTFT · 95.3 bench
1M
max out 128K
mistral-la-plateforme
共识领跑Unverified
22
@google·google/gemini-3-pro
49.5
100% conf
$2.00
47.9 lift/$
118 tok/s
ref 330ms TTFT · 95.8 bench
2M
max out 66K
google-vertex
稳健基线Unverified
2349.2
100% conf
$2.70
35.4 lift/$
162 tok/s
ref 175ms TTFT · 95.6 bench
1.0M
max out 66K
google-vertex
稳健基线Unverified
2446.4
100% conf
$3.60
27.3 lift/$
114 tok/s
ref 280ms TTFT · 98.2 bench
1.0M
max out 66K
google-vertex
稳健基线Unverified
2546.3
100% conf
$3.60
27.2 lift/$
108 tok/s
ref 265ms TTFT · 97.9 bench
1.1M
max out 128K
openai-direct
稳健基线Unverified
2646.2
100% conf
$3.60
27.1 lift/$
106 tok/s
ref 255ms TTFT · 97.7 bench
1M
max out 128K
anthropic-direct
稳健基线Unverified
27
@openai·openai/gpt-6-2-astra
32.2
100% conf
$14.40
6.9 lift/$
44 tok/s
ref 680ms TTFT · 99.1 bench
1.1M
max out 128K
openai-direct
延迟退化预警Unverified
28
@anthropic·anthropic/claude-fable-5-2
31.5
100% conf
$14.40
6.9 lift/$
42 tok/s
ref 710ms TTFT · 99.0 bench
1M
max out 128K
anthropic-direct
延迟退化预警Unverified