Skip to main content
benchlm.ai

Vals-hosted GPQA Diamond mirror (Vals GPQA Diamond mirror)

Vals AI hosted GPQA Diamond view with few-shot and zero-shot chain-of-thought task splits.

How BenchLM shows Vals GPQA Diamond mirror

BenchLM mirrors the public Vals AI Vals GPQA Diamond mirror leaderboard captured from https://www.vals.ai/benchmarks/gpqa and updated by Vals on May 16, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

Vals GPQA Diamond mirror is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

105 Vals rows3 task viewspublic datasetTasks: Overall, Few-Shot CoT, Zero-Shot CoTDisplay only
Vals GPQA Diamond mirror on Vals AIVals methodologyVals home

Vals GPQA Diamond score on Vals GPQA Diamond mirror — May 16, 2026

BenchLM mirrors the published vals gpqa diamond score view for Vals GPQA Diamond mirror. Gemini 3.1 Pro Preview leads the public snapshot at 95.45% , followed by GPT-5.5 (93.18%) and Gemini 3.5 Flash (92.68%). BenchLM does not use these results to rank models overall.

1

Gemini 3.1 Pro Preview

Google

google/gemini-3.1-pro-preview

95.45%
Overall —
2

GPT-5.5

OpenAI

openai/gpt-5.5

93.18%
Overall —
3

Gemini 3.5 Flash

Google

google/gemini-3.5-flash

92.68%
Overall —
105 modelsExternal benchmark mirrorsCurrentDisplay onlyUpdated May 16, 2026

The published Vals GPQA Diamond mirror snapshot is tightly clustered at the top: Gemini 3.1 Pro Preview sits at 95.45%, while the third row is only 2.78 points behind. The broader top-10 spread is 5.81 points, so many of the published scores sit in a relatively narrow band.

105 models have been evaluated on Vals GPQA Diamond mirror. The benchmark falls in the External benchmark mirrors category. BenchLM tracks this category separately from its weighted global scoring system, so these results are best compared on the dedicated Korean benchmark views. Vals GPQA Diamond mirror is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Vals GPQA Diamond mirror

Year

2026

Tasks

GPQA Diamond task splits

Format

Accuracy score

Difficulty

Graduate science reasoning

BenchLM keeps this Vals-hosted GPQA Diamond table separate from canonical GPQA source records.

Vals GPQA DiamondPublic benchmark source

BenchLM freshness & provenance

Version

Vals GPQA Diamond mirror 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Vals GPQA Diamond score table (105 models)

1
Gemini 3.1 Pro Previewgoogle/gemini-3.1-pro-preview
95.45%
2
GPT-5.5openai/gpt-5.5
93.18%
3
Gemini 3.5 Flashgoogle/gemini-3.5-flash
92.68%
4
Gemini 3 Pro Previewgoogle/gemini-3-pro-preview
91.67%
5
GPT-5.2openai/gpt-5.2-2025-12-11
91.67%
6
GPT-5.4openai/gpt-5.4-2026-03-05
91.67%
7
Grok 4.3grok/grok-4.3
91.41%
8
Claude Opus 4.7anthropic/claude-opus-4-7
90.15%
9
Muse Sparkmeta/muse_spark
89.65%
10
Claude Opus 4.6 Thinkinganthropic/claude-opus-4-6-thinking
89.65%
11
DeepSeek V4 Prodeepseek/deepseek-v4-pro
89.39%
12
Kimi K2.6 Thinkingkimi/kimi-k2.6-thinking
89.14%
13
Grok 4.20 0309 Reasoninggrok/grok-4.20-0309-reasoning
88.64%
14
Grok 4 0709grok/grok-4-0709
88.13%
15
Gemini 3 Flash Previewgoogle/gemini-3-flash-preview
87.88%
16
Qwen3.5 Plus Thinkingalibaba/qwen3.5-plus-thinking
87.37%
17
Qwen3.6 Plusalibaba/qwen3.6-plus
87.37%
18
GPT-5.1openai/gpt-5.1-2025-11-13
86.62%
19
MiniMax M2.7minimax/MiniMax-M2.7
86.62%
20
Claude Opus 4.5 20251101 Thinkinganthropic/claude-opus-4-5-20251101-thinking
85.86%
21
GPT-5openai/gpt-5-2025-08-07
85.61%
22
Claude Sonnet 4.6anthropic/claude-sonnet-4-6
85.61%
23
Grok 4 Fast Reasoninggrok/grok-4-fast-reasoning
85.35%
24
Qwen3 Maxalibaba/qwen3-max-2026-01-23
84.85%
25
GLM 5.1 Thinkingzai/glm-5.1-thinking
84.52%
26
Grok 4.1 Fast Reasoninggrok/grok-4-1-fast-reasoning
84.34%
27
O3openai/o3-2025-04-16
84.09%
28
Kimi K2.5 Thinkingkimi/kimi-k2.5-thinking
84.09%
29
GLM 5 Thinkingzai/glm-5-thinking
83.33%
30
GPT-5.4 Miniopenai/gpt-5.4-mini-2026-03-17
83.08%
31
Qwen3.5 Flashalibaba/qwen3.5-flash
82.83%
32
MiniMax M2.5 Lightningminimax/MiniMax-M2.5-Lightning
82.07%
33
Claude Sonnet 4.5 20250929 Thinkinganthropic/claude-sonnet-4-5-20250929-thinking
81.63%
34
Gemini 2.5 Flash Preview 09 2025google/gemini-2.5-flash-preview-09-2025
81.57%
35
Gemini 3.1 Flash Lite Previewgoogle/gemini-3.1-flash-lite-preview
81.06%
36
Gemini 2.5 Pro Exp 03 25google/gemini-2.5-pro-exp-03-25
80.81%
37
GPT-5 Miniopenai/gpt-5-mini-2025-08-07
80.30%
38
DeepSeek V3p2 Thinkingfireworks/deepseek-v3p2-thinking
80.30%
39
GLM 4.7zai/glm-4.7
80.05%
40
Claude Opus 4.5anthropic/claude-opus-4-5-20251101
79.55%
41
Qwen3 Maxalibaba/qwen3-max
79.55%
42
Grok 3 Mini Fast High Reasoninggrok/grok-3-mini-fast-high-reasoning
79.29%
43
GPT Oss 120bfireworks/gpt-oss-120b
78.54%
44
MiniMax M2.1minimax/MiniMax-M2.1
78.54%
45
Kimi K2 Thinkingkimi/kimi-k2-thinking
78.54%
46
Qwen3 Max Previewalibaba/qwen3-max-preview
77.78%
47
GPT-5.4 Nanoopenai/gpt-5.4-nano-2026-03-17
77.53%
48
Gemini 2.5 Flash Preview 09 2025 Thinkinggoogle/gemini-2.5-flash-preview-09-2025-thinking
76.52%
49
Claude Opus 4.1 20250805 Thinkinganthropic/claude-opus-4-1-20250805-thinking
76.26%
50
DeepSeek V3p2fireworks/deepseek-v3p2
76.26%
51
O3 Miniopenai/o3-mini-2025-01-31
75.50%
52
Claude 3.7 Sonnet 20250219 Thinkinganthropic/claude-3-7-sonnet-20250219-thinking
75.25%
53
Claude Sonnet 4 20250514 Thinkinganthropic/claude-sonnet-4-20250514-thinking
75.00%
54
O4 Miniopenai/o4-mini-2025-04-16
74.50%
55
GLM 4.6zai/glm-4.6
74.50%
56
Grok 3grok/grok-3
74.24%
57
O1openai/o1-2024-12-17
73.23%
58
Grok 3 Mini Fast Low Reasoninggrok/grok-3-mini-fast-low-reasoning
72.98%
59
Claude Haiku 4.5 20251001 Thinkinganthropic/claude-haiku-4-5-20251001-thinking
72.22%
60
GLM 4.5zai/glm-4.5
72.22%
61
Claude Opus 4anthropic/claude-opus-4-20250514
71.72%
62
Moonshotai Kimi K2 Instructtogether/moonshotai/Kimi-K2-Instruct
71.46%
63
Gemini 2.5 Flash Lite Preview 09 2025 Thinkinggoogle/gemini-2.5-flash-lite-preview-09-2025-thinking
70.20%
64
Qwen3 235b A22bfireworks/qwen3-235b-a22b
70.20%
65
Claude Opus 4.1anthropic/claude-opus-4-1-20250805
69.95%
66
Llama4 Maverick Instruct Basicfireworks/llama4-maverick-instruct-basic
69.44%
67
Claude Sonnet 4anthropic/claude-sonnet-4-20250514
69.44%
68
GPT Oss 20bfireworks/gpt-oss-20b
68.94%
69
Mistral Large 2512mistralai/mistral-large-2512
68.43%
70
GPT-4.1 Miniopenai/gpt-4.1-mini-2025-04-14
67.93%
71
Claude 3.7 Sonnetanthropic/claude-3-7-sonnet-20250219
67.42%
72
GPT-4.1openai/gpt-4.1-2025-04-14
65.40%
73
Gemini 2.0 Flash 001google/gemini-2.0-flash-001
65.15%
74
Grok 4.1 Fast Non Reasoninggrok/grok-4-1-fast-non-reasoning
65.15%
75
Gemini 2.5 Flash Lite Preview 09 2025google/gemini-2.5-flash-lite-preview-09-2025
64.14%
76
GPT-5 Nanoopenai/gpt-5-nano-2025-08-07
63.38%
77
Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561 Thinkingtogether/langston/nim/nvidia/llama-3.3-nemotron-super-49b-v1-42e84561-thinking
62.37%
78
Magistral Medium 2509mistralai/magistral-medium-2509
62.37%
79
Grok 4 Fast Non Reasoninggrok/grok-4-fast-non-reasoning
62.12%
80
DeepSeek V3 0324fireworks/deepseek-v3-0324
61.62%
81
Claude 3.5 Sonnetanthropic/claude-3-5-sonnet-20241022
59.34%
82
Mimo V2 Flashxiaomi/mimo-v2-flash
59.34%
83
Gemini 1.5 Pro 002google/gemini-1.5-pro-002
58.33%
84
Magistral Small 2509mistralai/magistral-small-2509
58.33%
85
Gemini 2.5 Flash Preview 04 17google/gemini-2.5-flash-preview-04-17
57.58%
86
DeepSeek V3fireworks/deepseek-v3
54.55%
87
GPT-4oopenai/gpt-4o-2024-11-20
53.79%
88
GPT-4.1 Nanoopenai/gpt-4.1-nano-2025-04-14
50.76%
89
Mistral Small 2402mistralai/mistral-small-2402
50.76%
90
Grok 2 1212grok/grok-2-1212
50.76%
91
GPT-4oopenai/gpt-4o-2024-05-13
50.25%
92
Meta Llama Llama 3.3 70B Instruct Turbotogether/meta-llama/Llama-3.3-70B-Instruct-Turbo
50.00%
93
Command A 03 2025cohere/command-a-03-2025
48.48%
94
Mistral Large 2411mistralai/mistral-large-2411
47.73%
95
Langston Nim Nvidia Llama 3.3 Nemotron Super 49b V1 42e84561together/langston/nim/nvidia/llama-3.3-nemotron-super-49b-v1-42e84561
47.73%
96
Meta Llama Llama 4 Scout 17B 16E Instructtogether/meta-llama/Llama-4-Scout-17B-16E-Instruct
46.97%
97
Gemini 1.5 Flash 002google/gemini-1.5-flash-002
45.96%
98
Mistral Small 2503mistralai/mistral-small-2503
44.19%
99
GPT-4o Miniopenai/gpt-4o-mini-2024-07-18
44.19%
100
Claude 3.5 Haikuanthropic/claude-3-5-haiku-20241022
37.88%
101
Jamba Large 1.6ai21labs/jamba-large-1.6
36.11%
102
Mistral Medium 3.5mistralai/mistral-medium-3.5
34.85%
103
Jamba Mini 1.6ai21labs/jamba-mini-1.6
32.07%
104
Command R Pluscohere/command-r-plus
31.06%
105
GPT-3.5 Turboopenai/gpt-3.5-turbo
30.56%

FAQ

What does Vals GPQA Diamond mirror measure?

Vals AI hosted GPQA Diamond view with few-shot and zero-shot chain-of-thought task splits.

Which model leads the published Vals GPQA Diamond mirror snapshot?

Gemini 3.1 Pro Preview currently leads the published Vals GPQA Diamond mirror snapshot with 95.45% vals gpqa diamond score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on Vals GPQA Diamond mirror?

105 AI models are included in BenchLM's mirrored Vals GPQA Diamond mirror snapshot, based on the public leaderboard captured on May 16, 2026.

Last updated: May 16, 2026 · mirrored from the public benchmark leaderboard

The AI models change fast. We track them for you.

For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.