DAYJOB: Law Can AI do a lawyer’s job?

Professional Multimodal Reasoning

GDP.pdf

Can frontier models master the documents that run the world? GDP.pdf is a multimodal and reasoning benchmark that takes real-world prompts and PDFs pulled directly from expert professional workflows.

RL Environments and the Hierarchy of Agentic Capabilities
Our RL environment run on 9 models revealed the core capabilities all agents need to master: tool use, planning, adaptability, groundedness, and common sense.
GDP.pdf Leaderboard
1
Gemini
4 Carbon (High reasoning)
56.4
%
1
Muse
Wonder 1.0 (Max reasoning)
54.8
%
1
Gemini
4 Hydrogen (xHigh reasoning)
54.2
%
1
Gemini
4 Argon (High reasoning)
53.6
%
1
Gemini
4 Neon (High reasoning)
53.4
%
1
Muse
Spark 1.3 (Max reasoning)
53
%
1
Claude
Fable 5.2 (Adaptive/Max)
51.8
%
1
GPT
6.1 Sol (Max reasoning)
51.6
%
1
Claude
Sonnet 5.5 (Adaptive/Max)
51.2
%
1
GPT
6 Astra (Max reasoning)
50.4
%
1
Claude
Fable 5.1 (Adaptive/Max)
49.4
%
1
GPT
6.2 Astra (Max reasoning)
49.2
%
1
Muse
Ember 1.0 (Max reasoning)
49.2
%
1
Claude
Opus 5.5 (Adaptive/Max)
48.8
%
1
GPT
6.2 Sol (Max reasoning)
48.8
%
1
Grok
4.7 (xHigh reasoning)
48.8
%
1
Kimi
K3.1 Pro (Max reasoning)
46.6
%
1
Gemini
3.8 Flash (High reasoning)
44.2
%
1
Hy
Hy4 (High reasoning)
43.6
%
1
Inkling
2 Small (xHigh reasoning)
43.6
%
1
Qwen
4 Max (xHigh reasoning)
43.6
%
1
Qwen
3.8 Max (xHigh reasoning)
42.8
%
1
Kimi
K3 (Max reasoning)
39.8
%
1
GLM
5.3 (Max reasoning)
36.6
%
1
Hy
Hy4 Preview (High reasoning)
35
%
1
Mistral
Large 4 (High reasoning)
32.8
%
1
MiniMax
M3.1 Flash (Max reasoning)
27
%

Stay Posted