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.
Enterprise Agents in Realistic RL Environments
EnterpriseBench: CoreCraft Agents
Stop testing models in tiny, self-contained environments. We built CoreCraft, a large-scale startup world, and deployed AI agents to solve real tasks. Our goal: to move agents beyond the cleanliness of the lab and into the chaos of enterprise reality.
EnterpriseBench: CoreCraft Agents Leaderboard
1
Claude
Fable 5.1 (Adaptive/Max)
77.4
%
1
Claude
Opus 5.5 (Adaptive/Max)
74.4
%
1
Claude
Fable 5 (Adaptive/Max)
70.3
%
1
Claude
Opus 5 (Adaptive/Max)
68.7
%
1
Grok
4.7 (xHigh reasoning)
68.2
%
1
Grok
4.6 (xHigh reasoning)
65.6
%
1
Hy
Hy4 Preview (High reasoning)
64.6
%
1
Muse
Spark 1.3 (xHigh reasoning)
61.9
%
1
Gemini
3.8 Flash (High reasoning)
58.5
%
1
Qwen
3.8 Flash (xHigh reasoning)
57.4
%
1
Gemini
3.7 Flash (High reasoning)
55.9
%
1
DeepSeek
V4 Pro (Max reasoning)
53.3
%
1
Claude
Opus 4.8 (Adaptive/Max)
52.3
%
1
GLM
5.3 (Max reasoning)
51.8
%
1
Muse
Glimmer 30B (xHigh reasoning)
51.8
%
1
GPT
5.5 (xHigh reasoning)
51.3
%
1
Claude
Sonnet 5 (Adaptive/Max)
51.3
%
1
Gemini
3.5 Flash (High reasoning)
50.8
%
1
GLM
5.3 Flash (Max reasoning)
46.7
%
1
GPT
5.6 Sol (Max reasoning)
46.2
%
1
DeepSeek
V4 Flash Vision (experimental) (Max reasoning)
45.6
%
1
GPT
6 Sol (Max reasoning)
43.1
%
1
GPT
6 Luna (Max reasoning)
43.1
%
1
Kimi
K3 (Max reasoning)
43.1
%
1
GPT
6 Astra (Max reasoning)
42.6
%
1
GPT
5.2 (xHigh reasoning)
42.6
%
1
GLM
5.2 (xHigh reasoning)
42.1
%
1
DeepSeek
V4.1 Flash (Max reasoning)
41.5
%
1
Hy
Hy3 (High reasoning)
41.0
%
1
Gemini
3.6 Flash (High reasoning)
39.5
%
1
GPT
5.4 (xHigh reasoning)
36.4
%
1
Claude
Opus 4.7 (Adaptive/Max)
35.9
%
1
DeepSeek
V4 Flash (High reasoning)
30.8
%
1
Claude
Opus 4.6 (Adaptive/Max)
30.8
%
1
Gemini
3.1 Pro (High reasoning)
27.2
%
1
Qwen
3.7 Max (Thinking on)
26.2
%
1
Kimi
K2.6 (Thinking on)
24.6
%
1
DeepSeek
V3.2 (High reasoning)
24.1
%
1
Grok
4.1 (Fast)
20.5
%
1
GPT
5.2 Codex (xHigh reasoning)
20.1
%
1
Gemini
3 Flash (High reasoning)
20.0
%
1
GLM
5 (Auto reasoning)
17.4
%
1
Claude
Sonnet 4.6 (High reasoning)
16.4
%
1
Gemini
3 Pro (High reasoning)
14.4
%
1
Qwen
3.5 Plus (Thinking on)
11.3
%
1
Nemotron
Lightning 3.5 30B A3B (Thinking on)
10.3
%
1
Nova
2 Pro (High reasoning)
8.9
%
1
Kimi
K2.5 (Thinking on)
8.7
%
1
Qwen
3 Max (Thinking on)
3.6
%
1
Mistral
Large 3
3.6
%
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
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.
Rank
Model
Score
23.9
%
Claude Opus 5.5 (Adaptive/Max)
Anthropic
21.5
%
GPT 6 Astra (Max)
OpenAI
20.3
%
Gemini 4 Argon (High)
Google
19.8
%
Claude Fable 5.1 (Adaptive/Max)
Anthropic
14.8
%
Muse Spark 1.3 (Max)
Meta
14.5
%
Grok 4.7 (xHigh)
xAI