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.
Antidote / Everyday
Antidote: Everyday Edition
Some leaderboards reward the answers that convince you in two seconds. Antidote rewards the answer that leaves you better off a month later. It's our real-world AI leaderboard, graded by doctors, lawyers, and engineers who read every word, check every citation, and run every line of code.
Antidote: Everyday Edition Leaderboard
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
1
Claude
Fable 5 (Adaptive/High)
1099
1085
-
1112
2
Gemini
3.1 Pro (High reasoning)
1091
1081
-
1101
3
Gemini
3.8 Flash (Medium reasoning)
1087
1070
-
1104
4
Gemini
3.6 Flash (Medium reasoning)
1083
1069
-
1097
5
Gemini
3.5 Flash (Medium reasoning)
1079
1068
-
1090
6
Kimi
K3 (Max reasoning)
1077
1063
-
1091
7
Gemini
3.7 Flash (Medium reasoning)
1075
1059
-
1090
8
GLM
5.3 (Max reasoning)
1067
1051
-
1083
9
Claude
Fable 5.1 (Adaptive/High)
1065
1048
-
1083
10
Claude
Opus 5 (Adaptive/High)
1055
1041
-
1069
11
Qwen
3.7 Max (Thinking on)
1051
1040
-
1061
12
Kimi
K2.6 (Thinking on)
1045
1036
-
1054
13
Claude
Opus 4.7 (High reasoning)
1042
1031
-
1053
13
Claude
Opus 4.6 (High reasoning)
1042
1032
-
1052
15
GLM
5.2 (Auto reasoning)
1040
1027
-
1052
16
Claude
Opus 4.8 (High reasoning)
1030
1018
-
1043
17
Muse
Spark 1.2 (Auto reasoning)
1028
1011
-
1046
18
GPT
5.6 Sol (Medium reasoning)
1027
1014
-
1041
19
Hy
Hy3 (No reasoning)
1022
1007
-
1037
20
DeepSeek
V4 Pro (High reasoning)
1018
1003
-
1033
21
DeepSeek
V4 Flash Vision (experimental)
1015
999
-
1030
22
Muse
Spark 1.1 (Medium reasoning)
1014
998
-
1030
23
GPT
5.5 (Medium reasoning)
1013
1000
-
1025
24
Claude
Sonnet 4.6 (High reasoning)
1012
1000
-
1024
25
DeepSeek
V4 Pro (preview) (High reasoning)
1007
995
-
1019
26
Kimi
K2.5 (Thinking on)
1006
995
-
1018
27
Grok
4.20 Beta
1001
989
-
1012
28
Qwen
3.5 Plus (Thinking on)
999
987
-
1011
29
Hy
Hy4 Preview (High reasoning)
992
974
-
1009
30
Gemini
3.5 Flash-Lite (Minimal reasoning)
991
977
-
1005
30
Inkling
Inkling (High reasoning)
991
977
-
1005
32
Qwen
3.8 Max (xHigh reasoning)
985
970
-
1000
33
Grok
4.5 (High reasoning)
973
960
-
986
34
DeepSeek
V4 Flash (preview) (High reasoning)
966
954
-
978
34
Grok
4.6 (High reasoning)
966
950
-
981
36
DeepSeek
V3.2 (No reasoning)
954
940
-
967
37
Grok
4.3 (default)
950
939
-
960
38
Mistral
Large 3
947
938
-
957
38
Ernie
5.1
947
936
-
958
40
Qwen
3.8 Flash (xHigh reasoning)
944
927
-
962
41
Claude
Haiku 4.5
932
919
-
944
42
Muse
Glimmer 30B (High reasoning)
931
915
-
947
43
GPT
5.4 Mini (No reasoning)
929
917
-
941
44
Gemma
3 12B
893
880
-
905
45
Nemotron
Lightning 3.5 30B A3B (Thinking on)
888
871
-
904
46
Ernie
4.5 300B
863
850
-
876
47
Nova
2 Pro (No reasoning)
819
809
-
830
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