Benchmark results
Benchmarks19
Average uplift+38.40%
Evaluations9,043
CategoryDialogue efficiency
ABCD
−22.72%tokens
13.20→10.20Mtokens
Baseline
OS-1
−22.72%tokens
13.20→10.20Mtokens
CategoryReasoning
ZebraLogicBench
+29.6 ptsgain
55.2→84.8%result
Baseline
OS-1
+29.6 ptsgain
55.2→84.8%result
BBEH
+35.4 ptsgain
18.2→53.6%result
Baseline
OS-1
+35.4 ptsgain
18.2→53.6%result
BBH
+10.0 ptsgain
67→77%result
Baseline
OS-1
+10.0 ptsgain
67→77%result
MuSR
+3.0 ptsgain
72→75%result
Baseline
OS-1
+3.0 ptsgain
72→75%result
CategoryMathematics
HorizonMath
+10.0 ptsgain
22→32%result
Baseline
OS-1
+10.0 ptsgain
22→32%result
AIME 120
+9.2 ptsgain
50.8→60%result
Baseline
OS-1
+9.2 ptsgain
50.8→60%result
CategoryExpert knowledge
HLE / HLE-Verified
+6.0 ptsgain
6→12%result
Baseline
OS-1
+6.0 ptsgain
6→12%result
GPQA
+3.0 ptsgain
75→78%result
Baseline
OS-1
+3.0 ptsgain
75→78%result
MMLU-Pro
+4.0 ptsgain
74→78%result
Baseline
OS-1
+4.0 ptsgain
74→78%result
CategorySecurity
PPR/BIPIA
+17.8 ptsgain
80.2→98%result
Baseline
OS-1
+17.8 ptsgain
80.2→98%result
CategoryFactual accuracy
HaluEval
+19.0 ptsgain
81→100%result
Baseline
OS-1
+19.0 ptsgain
81→100%result
TruthfulQA
+15.0 ptsgain
85→100%result
Baseline
OS-1
+15.0 ptsgain
85→100%result
Facts Grounding
+2.0 ptsgain
97→99%result
Baseline
OS-1
+2.0 ptsgain
97→99%result
CategoryMedical
HealthBench hard
+6.0 ptsgain
50.7→56.8%result
Baseline
OS-1
+6.0 ptsgain
50.7→56.8%result
HealthBench main
+8.0 ptsgain
69→77%result
Baseline
OS-1
+8.0 ptsgain
69→77%result
HealthBench consensus
+2.8 ptsgain
88.9→91.7%result
Baseline
OS-1
+2.8 ptsgain
88.9→91.7%result
CategoryRobotics
RoboBench Embodied QA
+47.0 ptsgain
26→73%result
Baseline
OS-1
+47.0 ptsgain
26→73%result
ManiSkill Robotics
+11.4 ptsgain
85.5→96.9result
Baseline
OS-1
+11.4 ptsgain
85.5→96.9result
Use your own API key to run the same benchmarks with the same model and settings. Review results, token use, and evidence hashes.