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.
Agentic Finance
DAYJOB: Finance
DAYJOB: Finance measures long-horizon finance agents across corporate finance, banking, credit, investing, and real assets. It's built around the messy reality of knowledge work: realistic requests, complex environments, and end-to-end professional judgment.
DAYJOB: Finance Leaderboard
1
Claude
Opus 5.5 (Adaptive/Max)
23.9
%
1
GPT
6 Astra (Max reasoning)
21.5
%
1
Gemini
4 Argon (High reasoning)
20.3
%
1
Claude
Fable 5.1 (Adaptive/Max)
19.8
%
1
Muse
Spark 1.3 (Max reasoning)
14.8
%
1
Grok
4.7 (xHigh reasoning)
14.5
%
1
Grok
4.6 (xHigh reasoning)
12.8
%
1
Claude
Opus 5 (Adaptive/Max)
11.3
%
1
GPT
6 Sol (Max reasoning)
9.3
%
1
Claude
Fable 5 (Adaptive/Max)
9
%
1
Hy
Hy4 Preview (High reasoning)
7.3
%
1
GLM
5.3 (Max reasoning)
6
%
1
GPT
5.6 Sol (Max reasoning)
5.5
%
1
Kimi
K3 (Max reasoning)
3.8
%
1
Qwen
3.8 Max (xHigh reasoning)
3
%
1
Gemini
3.7 Flash (High reasoning)
2.8
%
1
Gemini
3.8 Flash (High reasoning)
2.8
%
1
GLM
5.3 Flash (Max reasoning)
2.3
%
1
Muse
Spark 1.2 (xHigh reasoning)
2.3
%
1
Claude
Sonnet 5 (Adaptive/Max)
2.3
%
1
GPT
6 Luna (Max reasoning)
2.3
%
1
GPT
5.6 Luna (Max reasoning)
1.8
%
1
GPT
5.6 Terra (xHigh reasoning)
1.8
%
1
DeepSeek
V4 Pro (Max reasoning)
1
%
1
Hy
Hy3 (High reasoning)
0.3
%
1
DeepSeek
V4 Flash (Max reasoning)
0
%
1
Gemini
3.1 Pro (High reasoning)
0
%
1
Inkling
Inkling (Max reasoning)
0
%
1
Kimi
K2.7 Code (Max reasoning)
0
%
1
Mistral
Large 3
0
%
1
Muse
Glimmer 30B (xHigh reasoning)
0
%
1
Nemotron
3 Ultra
0
%
VIEW ALL BENCHMARKS
Agentic Coding
sudo L7
Coding agents are getting very good at being L3s. But do they have the judgment to become L7s? Across 60 tasks, agents must navigate unfamiliar systems, choose architectures, spot hidden risks, and know when technically correct code is still the wrong thing to ship.
Rank
Model
Score
41.7
%
Claude Sonnet 5.5 (Adaptive/Max)
Anthropic
41.7
%
Claude Opus 5.5 (Adaptive/Max)
Anthropic
33.9
%
GPT 6 Astra (Max)
OpenAI
31.1
%
Claude Fable 5.1 (Adaptive/Max)
Anthropic
28.9
%
GPT 6.1 Sol (Max)
OpenAI
20.6
%
Qwen 3.8 Max
Alibaba Cloud
Enterprise Instruction Following
ComplexConstraints
A benchmark for professional instruction following, where constraints depend on each other, fire conditionally, and must be inferred from context.
Rank
Model
Score
GPT 6 Astra (Max)
57.7
%
OpenAI
Gemini 4 Argon (High)
53.3
%
Google
Muse Spark 1.3 (xHigh)
51.9
%
Meta
GPT 6 Sol (Max)
50.8
%
OpenAI
GPT 5.6 Sol (Max)
50.5
%
OpenAI
Claude Opus 5.5 (Adaptive/Max)
49.9
%
Anthropic