live
weekly refresh
basedagi.org
benchmark evidence

UGI — ugi-overall

UGI Leaderboard — ugi-overall score from DontPlanToEnd/UGI-Leaderboard

winner on UGI — ugi-overall
direct benchmark result, not a broad vertical composite | source row dated 2025-09-30
scored on 2026-04-25 · stale source data (35d)
latest mapped results | top 20
#ModelScoreEvidenceTested
1Mistral Large
Mistralai
59.2
model-only
independent_benchmark
2025-09-30
2MoonshotAI: Kimi K2 Thinking
Moonshotai
52.7
model-only
independent_benchmark
2025-11-07
3OpenAI: o3
Openai
52.1
model-only
independent_benchmark
2025-09-24
4OpenAI: GPT-5.5
Openai
51.2
model-only
independent_benchmark
2026-04-25
5DeepSeek: R1
Deepseek
51.0
model-only
independent_benchmark
2025-09-21
6Google: Gemini 2.5 Pro
Google
47.8
model-only
independent_benchmark
2025-09-09
7OpenAI: GPT-4.1
Openai
47.5
model-only
independent_benchmark
2025-09-24
8OpenAI: GPT-5.4
Openai
47.0
model-only
independent_benchmark
2026-03-05
9OpenAI: GPT-5.1
Openai
44.3
model-only
independent_benchmark
2025-11-16
10OpenAI: o1
Openai
42.3
model-only
independent_benchmark
2025-09-24
11OpenAI: GPT-5
Openai
42.1
model-only
independent_benchmark
2025-09-24
12OpenAI: o4 Mini
Openai
38.2
model-only
independent_benchmark
2025-10-04
13Meta: Llama 3.3 70B Instruct
Meta Llama
37.9
model-only
independent_benchmark
2025-09-10
14Qwen: Qwen3 235B A22B
Qwen
35.2
model-only
independent_benchmark
2025-09-21
15Mistral: Mistral Nemo
Mistralai
33.4
model-only
independent_benchmark
2025-09-06
16Qwen: Qwen3 32B
Qwen
31.0
model-only
independent_benchmark
2025-09-13
17OpenAI: GPT-5.2
Openai
31.0
model-only
independent_benchmark
2026-03-04
18Qwen: Qwen3 30B A3B
Qwen
28.8
model-only
independent_benchmark
2025-11-08
19Anthropic: Claude Sonnet 4
Anthropic
23.2
model-only
independent_benchmark
2025-09-12
20Google: Gemma 3 27B
Google
20.4
model-only
independent_benchmark
2025-09-06
what this result means

UGI Leaderboard — ugi-overall score from DontPlanToEnd/UGI-Leaderboard

This benchmark contributes direct public evidence. Read its scope before generalizing the result.

A win here is a win on UGI — ugi-overall. Broad task pages require independent corroboration before naming a general winner.

source record
category: uncensored
metric: accuracy
matched models: 20
latest source date: 2026-04-25
direction: higher is better
inspect upstream source ->