Hugging Face's logo

Datasets:
regolo
/
brick-skill-tables

Dataset card Files
xet
 Community
2
massaindustries commited on
Commit
7cbbfb3
·
verified ·
1 Parent(s): a4c7834

Add/update skill card: gpt-5.6-sol

Browse files
Files changed (2) hide show
  1. gpt-5.6-sol.json +36 -0
  2. index.json +35 -29
gpt-5.6-sol.json ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": "gpt-5.6-sol",
3
+ "provider": "openai",
4
+ "capabilities": [
5
+ "coding",
6
+ "creative_synthesis",
7
+ "instruction_following",
8
+ "math_reasoning",
9
+ "planning_agentic",
10
+ "world_knowledge"
11
+ ],
12
+ "skill_vector": [
13
+ 0.9,
14
+ 0.91,
15
+ 0.86,
16
+ 0.94,
17
+ 0.92,
18
+ 0.94
19
+ ],
20
+ "source": "benchmark",
21
+ "confidence": [
22
+ "high",
23
+ "low",
24
+ "medium",
25
+ "medium",
26
+ "high",
27
+ "medium"
28
+ ],
29
+ "imputed_capabilities": [
30
+ "creative_synthesis"
31
+ ],
32
+ "support": null,
33
+ "subset_hash": null,
34
+ "date": "2026-07-12",
35
+ "notes": "Benchmark-derived skill vector for GPT-5.6 Sol (flagship tier, released 2026-07-09). OpenAI's launch suite is deliberately agentic and omits AIME/GPQA/MMLU/IFEval, so only coding and planning_agentic have direct public benchmarks (confidence high); math/world/instruction are anchored to the composite intelligence index (confidence medium); creative_synthesis is imputed (confidence low). Sourcing per dimension: coding=Terminal-Bench 2.1 88.8% (SOTA) + Artificial Analysis Coding Agent Index 80/100 (SOTA); planning_agentic=Agents' Last Exam 53.6 (SOTA, +13 over Fable 5) + OSWorld 2.0 62.6% (SOTA); math_reasoning + world_knowledge=Artificial Analysis Intelligence Index v4.1 Sol=59 (near-top), no standalone AIME/GPQA published; instruction_following=no public IFEval, estimated from the composite index; creative_synthesis=imputed as the mean of the known coordinates. Sources: openai.com/index/gpt-5-6, artificialanalysis.ai/articles/gpt-5-6-has-landed, vellum.ai/blog/gpt-5-6-benchmarks-explained. A `measured` extraction on the frozen probe set overrides this. See README.md."
36
+ }
index.json CHANGED
@@ -7,8 +7,14 @@
7
  "planning_agentic",
8
  "world_knowledge"
9
  ],
10
- "date": "2026-06-11",
11
  "models": [
 
 
 
 
 
 
12
  {
13
  "model": "claude-opus-4-8",
14
  "provider": "anthropic",
@@ -22,63 +28,63 @@
22
  "source": "benchmark"
23
  },
24
  {
25
- "model": "claude-haiku-4-5",
26
- "provider": "anthropic",
27
- "file": "claude-haiku-4-5.json",
28
  "source": "benchmark"
29
  },
30
  {
31
- "model": "gpt-5.5",
32
- "provider": "openai",
33
- "file": "gpt-5.5.json",
34
  "source": "benchmark"
35
  },
36
  {
37
- "model": "gpt-5.4",
38
- "provider": "openai",
39
- "file": "gpt-5.4.json",
40
  "source": "benchmark"
41
  },
42
  {
43
- "model": "gpt-5.4-mini",
44
- "provider": "openai",
45
- "file": "gpt-5.4-mini.json",
46
  "source": "benchmark"
47
  },
48
  {
49
- "model": "o3",
50
  "provider": "openai",
51
- "file": "o3.json",
52
  "source": "benchmark"
53
  },
54
  {
55
- "model": "o3-mini",
56
  "provider": "openai",
57
- "file": "o3-mini.json",
58
  "source": "benchmark"
59
  },
60
  {
61
- "model": "gemini-3.1-pro",
62
- "provider": "google",
63
- "file": "gemini-3.1-pro.json",
64
  "source": "benchmark"
65
  },
66
  {
67
- "model": "gemini-3.5-flash",
68
- "provider": "google",
69
- "file": "gemini-3.5-flash.json",
70
  "source": "benchmark"
71
  },
72
  {
73
- "model": "gemini-3.1-flash-lite",
74
- "provider": "google",
75
- "file": "gemini-3.1-flash-lite.json",
76
  "source": "benchmark"
77
  },
78
  {
79
- "model": "gemini-2.5-pro",
80
- "provider": "google",
81
- "file": "gemini-2.5-pro.json",
82
  "source": "benchmark"
83
  }
84
  ]
 
7
  "planning_agentic",
8
  "world_knowledge"
9
  ],
10
+ "date": "2026-07-12",
11
  "models": [
12
+ {
13
+ "model": "claude-haiku-4-5",
14
+ "provider": "anthropic",
15
+ "file": "claude-haiku-4-5.json",
16
+ "source": "benchmark"
17
+ },
18
  {
19
  "model": "claude-opus-4-8",
20
  "provider": "anthropic",
 
28
  "source": "benchmark"
29
  },
30
  {
31
+ "model": "gemini-2.5-pro",
32
+ "provider": "google",
33
+ "file": "gemini-2.5-pro.json",
34
  "source": "benchmark"
35
  },
36
  {
37
+ "model": "gemini-3.1-flash-lite",
38
+ "provider": "google",
39
+ "file": "gemini-3.1-flash-lite.json",
40
  "source": "benchmark"
41
  },
42
  {
43
+ "model": "gemini-3.1-pro",
44
+ "provider": "google",
45
+ "file": "gemini-3.1-pro.json",
46
  "source": "benchmark"
47
  },
48
  {
49
+ "model": "gemini-3.5-flash",
50
+ "provider": "google",
51
+ "file": "gemini-3.5-flash.json",
52
  "source": "benchmark"
53
  },
54
  {
55
+ "model": "gpt-5.4",
56
  "provider": "openai",
57
+ "file": "gpt-5.4.json",
58
  "source": "benchmark"
59
  },
60
  {
61
+ "model": "gpt-5.4-mini",
62
  "provider": "openai",
63
+ "file": "gpt-5.4-mini.json",
64
  "source": "benchmark"
65
  },
66
  {
67
+ "model": "gpt-5.5",
68
+ "provider": "openai",
69
+ "file": "gpt-5.5.json",
70
  "source": "benchmark"
71
  },
72
  {
73
+ "model": "gpt-5.6-sol",
74
+ "provider": "openai",
75
+ "file": "gpt-5.6-sol.json",
76
  "source": "benchmark"
77
  },
78
  {
79
+ "model": "o3",
80
+ "provider": "openai",
81
+ "file": "o3.json",
82
  "source": "benchmark"
83
  },
84
  {
85
+ "model": "o3-mini",
86
+ "provider": "openai",
87
+ "file": "o3-mini.json",
88
  "source": "benchmark"
89
  }
90
  ]