Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Long-context reasoning on RULER 200k tokens
Loading...
100
Single-NIAH Score
Full KV
95
97.5
100
102.5
Apr 12, 2026
Single-NIAH Score
Multi-keys NIAH Score
QA Score
Updated 4d ago
Evaluation Results
Method
Method
Links
Single-NIAH Score
Multi-keys NIAH Score
QA Score
Full KV
Backbone=Qwen3-4B-Inst...
2026.04
100
92
95
IceCache
Backbone=Qwen3-4B-Inst...
2026.04
100
91.3
95
IceCache (r)
Backbone=Qwen3-4B-Inst...
2026.04
100
91
94.3
Feedback
Search any
task
Search any
task