U4 Benchmarks

Uhuru AI � U4

Benchmark Performance

U4 is evaluated against the world's top frontier models on standard AI benchmarks — and dominates every rival on African language, cultural and contextual tasks.

90.1%
Avg Standard Score
across 8 benchmarks
9
Best-in-Class Wins
of 14 total benchmarks
91.3%
African Context
+25pts avg lead over rivals
#2
Overall Ranking
global standard · #1 African AI
U4 ranks #2 globally on standard benchmarks — leading on commonsense reasoning, science, and truthfulness. Trails on math and code where GPT-5 and Claude Opus set the bar.
BenchmarkDescription
U4
Uhuru AI
ChatGPT / GPT-5
OpenAI
Claude Opus 4.6
Anthropic
Claude Sonnet 4.6
Anthropic
Gemini 2.5 Pro
Google
Perplexity
Perplexity AI
DeepSeek-V3
DeepSeek
MMLU
Massive Multitask Language Understanding — 57 academic subjects90.1%
94.2%BEST
91.3%
88.7%
89.8%
88.2%
88.5%
HumanEval
Code generation — Python function completion from docstrings84.6%
92%
95%BEST
91.2%
89.7%
78.4%
89.3%
GSM8K
Grade school math — multi-step arithmetic reasoning problems88%
96.8%BEST
96.4%
92.6%
91.4%
85.6%
92.8%
HellaSwag
Common-sense NLI — sentence completion requiring world knowledge91.7%BEST
87.4%
87.2%
85.8%
88.2%
89.1%
85.6%
ARC-Challenge
Science questions requiring reasoning beyond simple retrieval93.4%BEST
93.1%
92.4%
90.6%
90.7%
91.2%
90.2%
TruthfulQA
Resistance to producing false beliefs and common misconceptions89.3%BEST
85.4%
86.2%
84.7%
83.4%
87.8%
80.1%
MT-Bench
Multi-turn conversation quality judged by GPT-4 (score out of 10)8.7/10
9.1/10
9.2/10BEST
8.9/10
8.8/10
8.4/10
8.6/10
MATH
Competition-level math problems across algebra, geometry, calculus93.8%
97%BEST
96.2%
91%
95%
79.3%
92.1%

Uhuru scores measured via live API evaluation (March 2026). Competitor scores from public model cards. MT-Bench scored out of 10.