Thematic Generalisation

AI benchmark results and model performance

Type: numerical

View benchmark source

Recorded Results

Average Score

57.23

Scores Over Time

Individual benchmark scores plotted by date.

Models Using This Benchmark

Organisation	Model	Reported	Top Score	Info	Self Reported	Source
Mistral	Mistral Medium 3.1	12 Aug 2025	20.30	Top-1 accuracy 32.3%; cases 703	No	Source
SpaceXAI	Grok 4.20	17 Feb 2026	22	Non-reasoning; Top-1 accuracy 33.4%; cases 703	No	Source
Mistral	Mistral Large 3.0	02 Dec 2025	23	Top-1 accuracy 34.0%; cases 703	No	Source
OpenAI	GPT 5.4	05 Mar 2026	27.10	No reasoning; Top-1 accuracy 44.0%; cases 703	No	Source
MiniMax	MiniMax M2.7	18 Mar 2026	39.30	Top-1 accuracy 63.4%; cases 703	No	Source
Arcee AI	Trinity Large Thinking	01 Apr 2026	41.60	Top-1 accuracy 66.1%; cases 703	No	Source
Baidu	Ernie 5.0	22 Jan 2026	41.70	Top-1 accuracy 65.3%; cases 703	No	Source
Qwen	Qwen 3.5 27B	24 Feb 2026	45.50	Top-1 accuracy 71.3%; cases 703	No	Source
Xiaomi	MiMo V2 Pro	18 Mar 2026	45.90	Top-1 accuracy 68.8%; cases 703	No	Source
Qwen	Qwen 3.5 122B A10B	24 Feb 2026	51.20	Top-1 accuracy 76.5%; cases 703	No	Source
Anthropic	Claude Opus 4.7	16 Apr 2026	52.60	No reasoning; Top-1 accuracy 71.6%; cases 703	No	Source
Google	Gemma 4 31B	02 Apr 2026	53	Reasoning; Top-1 accuracy 76.1%; cases 703	No	Source
ByteDance	Seed 2.0 Pro	14 Feb 2026	57.10	Top-1 accuracy 77.0%; cases 703	No	Source
Qwen	Qwen 3.6 Plus	01 Apr 2026	59.50	Top-1 accuracy 81.5%; cases 703	No	Source
OpenAI	GPT 5.4 Mini	17 Mar 2026	61.70	xHigh reasoning; Top-1 accuracy 80.8%; cases 703	No	Source
Google	Gemini 3.1 Flash Lite Preview	03 Mar 2026	63.30	Top-1 accuracy 82.1%; cases 703	No	Source
DeepSeek	DeepSeek V3.2	01 Dec 2025	65	Top-1 accuracy 81.8%; cases 703	No	Source
Qwen	Qwen 3.5 397B A17B	16 Feb 2026	65.10	Top-1 accuracy 82.4%; cases 703	No	Source
Anthropic	Claude Sonnet 4.6	17 Feb 2026	68.50	No reasoning; Top-1 accuracy 82.5%; cases 703	No	Source
Anthropic	Claude Opus 4.6	05 Feb 2026	68.80	No reasoning; Top-1 accuracy 84.1%; cases 703	No	Source
Moonshot	Kimi K2.5	27 Jan 2026	69.40	Thinking; Top-1 accuracy 84.8%; cases 703	No	Source
z.AI	GLM 5.1	07 Apr 2026	69.80	Top-1 accuracy 85.9%; cases 703	No	Source
Google	Gemini 3.1 Pro Preview	19 Feb 2026	79.40	Top-1 accuracy 91.5%; cases 703	No	Source

Thematic Generalisation

Recorded Results

Average Score

Models Using This Benchmark

Score Range

Leading Model (lowest score)