benchmark
.
space
benchmarks
rankings
compare
voices
transcripts
papers
articles
GDPval-AA leaderboard
GDPval-AA
10 models tested · Updated 2026-07-21 · Verified sources only
Claude Fable 5
leads at
1747.0Elo
1
Claude Fable 5
Anthropic ·
Artificial Analysis
· 2026-07-21
#1 on GDPval-AA v2. Highest Elo on real-world professional work deliverables.
1747.0Elo
2
GPT-5.6 Sol
OpenAI ·
Artificial Analysis
· 2026-07-21
#2 on GDPval-AA v2, just 11 Elo behind Claude Fable 5.
1736.0Elo
3
Kimi K3
Moonshot ·
Artificial Analysis
· 2026-07-21
#4 on GDPval-AA v2. Strong showing for open-weight model.
1686.0Elo
4
Claude Sonnet 5
Anthropic ·
Artificial Analysis
· 2026-07-21
#6 on GDPval-AA v2.
1603.0Elo
5
Claude Opus 4.8
Anthropic ·
Artificial Analysis
· 2026-07-21
#7 on GDPval-AA v2.
1593.0Elo
6
Grok 4.5
xAI ·
Artificial Analysis
· 2026-07-21
#13 on GDPval-AA v2.
1527.0Elo
7
GLM-5.2
Z AI ·
Artificial Analysis
· 2026-07-21
Real workplace deliverables across 44 occupations. 372 Elo above Qwen 3.6 27B.
1510.0Elo
8
Gemini 3.6 Flash
Google ·
Artificial Analysis
· 2026-07-21
#22 on GDPval-AA v2. Strong for a Flash-tier model.
1423.0Elo
9
DeepSeek V4 Pro
DeepSeek ·
Artificial Analysis
· 2026-07-21
#34 on GDPval-AA v2.
1306.0Elo
10
Qwen 3.6 27B
Alibaba ·
Artificial Analysis
· 2026-07-21
Real workplace deliverables (44 occupations). 372 Elo below GLM-5.2.
1138.0Elo