Reasoning benchmark

NYT Connections (extended) leaderboard

As of October 2026, GPT-6 Astra has the highest published NYT Connections (extended) score on Noometry at 98.1%, out of 91 models with results.

Last verified October 10, 2026

About NYT Connections (extended)

New York Times Connections puzzles made harder with extra distractor words: sort 16 or more words into groups of four by a hidden theme.

Category: Reasoning
Introduced: 2024
Size: 940 puzzles
Format: Word grouping
Unit: Percent (random guessing ≈ 0%)
Official site: github.com

Top 15 models

GPT-6 Astra 98.1%
Gemini 3.1 Pro Preview 97.4%
Gemini 3.8 Flash 97.4%
GPT-5.5 96.2%
GPT-6.1 Sol 95.5%
Gemini 3 Pro 94.4%
Claude Opus 5 94.3%
Gemini 3.7 Flash 94%
GPT-5.6 Sol 93.8%
Kimi K3 93.6%
Claude Fable 5 92.7%
Gemini 3.5 Flash 92.6%
Claude Opus 4.6 92.1%
DeepSeek V4 Pro 91.3%
GPT-5.4 91.3%
9092949698100

Sponsored placements are available on pages like this one. Advertise on Noometry

All results

NYT Connections (extended) results by model
#	Model	Provider	Score	Setting	Source
1	GPT-6 Astra	OpenAI	98.1%	xhigh reasoning	Lech Mazur benchmarks
2	Gemini 3.1 Pro Preview	Google	97.4%		Lech Mazur benchmarks
3	Gemini 3.8 Flash	Google	97.4%	high reasoning	Lech Mazur benchmarks
4	GPT-5.5	OpenAI	96.2%	xhigh reasoning	Lech Mazur benchmarks
5	GPT-6.1 Sol	OpenAI	95.5%	high reasoning	Lech Mazur benchmarks
6	Gemini 3 Pro	Google	94.4%		Lech Mazur benchmarks
7	Claude Opus 5	Anthropic	94.3%	xhigh reasoning	Lech Mazur benchmarks
8	Gemini 3.7 Flash	Google	94%		Lech Mazur benchmarks
9	GPT-5.6 Sol	OpenAI	93.8%	xhigh reasoning	Lech Mazur benchmarks
10	Kimi K3	Moonshot AI	93.6%		Lech Mazur benchmarks
11	Claude Fable 5	Anthropic	92.7%	high reasoning	Lech Mazur benchmarks
12	Gemini 3.5 Flash	Google	92.6%		Lech Mazur benchmarks
13	Claude Opus 4.6	Anthropic	92.1%	high reasoning	Lech Mazur benchmarks
14	DeepSeek V4 Pro	DeepSeek	91.3%	high reasoning	Lech Mazur benchmarks
15	GPT-5.4	OpenAI	91.3%	xhigh reasoning	Lech Mazur benchmarks
16	Claude Opus 4.8	Anthropic	91.1%	xhigh reasoning	Lech Mazur benchmarks
17	GPT-6 Sol	OpenAI	90.1%	high reasoning	Lech Mazur benchmarks
18	Claude Fable 5.1	Anthropic	90%	high reasoning	Lech Mazur benchmarks
19	DeepSeek V4.1 Flash	DeepSeek	89.6%	high reasoning	Lech Mazur benchmarks
20	DeepSeek V4 Flash	DeepSeek	89.6%		Lech Mazur benchmarks
21	Grok 4.20 Multi-Agent	xAI	89.6%		Lech Mazur benchmarks
22	Gemini 3.6 Flash	Google	89%		Lech Mazur benchmarks
23	Claude Opus 5.5	Anthropic	88.5%	high reasoning	Lech Mazur benchmarks
24	Qwen3.8 Max	Alibaba (Qwen)	88.3%		Lech Mazur benchmarks
25	Grok 4.1 Fast	xAI	87.4%		Lech Mazur benchmarks
26	Kimi K2.6	Moonshot AI	87.2%		Lech Mazur benchmarks
27	Grok 4.20 (Non-Reasoning)	xAI	85.4%	reasoning	Lech Mazur benchmarks
28	Muse Spark 1.3	Meta	85.1%	high reasoning	Lech Mazur benchmarks
29	Qwen3.7 Max	Alibaba (Qwen)	85.1%		Lech Mazur benchmarks
30	Muse Spark 1.1	Meta	84.9%	high reasoning	Lech Mazur benchmarks
31	GPT-5.2	OpenAI	83.6%	xhigh reasoning	Lech Mazur benchmarks
32	Gemini 3 Flash Preview	Google	83.1%		Lech Mazur benchmarks
33	Claude Sonnet 4.6	Anthropic	80.9%	high reasoning	Lech Mazur benchmarks
34	Claude Sonnet 5.5	Anthropic	80.5%	high reasoning	Lech Mazur benchmarks
35	Grok 4.6	xAI	80%	xhigh reasoning	Lech Mazur benchmarks
36	Grok 4.5	xAI	79.9%	high reasoning	Lech Mazur benchmarks
37	GPT-5.2 Pro	OpenAI	79.3%		Lech Mazur benchmarks
38	Muse Spark 1.2	Meta	79.2%	high reasoning	Lech Mazur benchmarks
39	GPT-5.6 Terra	OpenAI	78.4%	high reasoning	Lech Mazur benchmarks
40	GLM-5.1	Z.ai (Zhipu)	77.7%		Lech Mazur benchmarks
41	Grok 4.7	xAI	76.8%	high reasoning	Lech Mazur benchmarks
42	Claude Sonnet 5	Anthropic	75.1%	high reasoning	Lech Mazur benchmarks
43	GLM-5	Z.ai (Zhipu)	74.8%		Lech Mazur benchmarks
44	Qwen3.7 Plus	Alibaba (Qwen)	74.8%		Lech Mazur benchmarks
45	GLM-5.2	Z.ai (Zhipu)	74.3%	high reasoning	Lech Mazur benchmarks
46	GLM-5.3	Z.ai (Zhipu)	74.2%	high reasoning	Lech Mazur benchmarks
47	Qwen3.6 Max Preview	Alibaba (Qwen)	74.1%		Lech Mazur benchmarks
48	Gemma 4 31B IT	Google	70.6%		Lech Mazur benchmarks
49	Kimi K2.5	Moonshot AI	69.9%		Lech Mazur benchmarks
50	GPT-5.6 Luna	OpenAI	69.4%	high reasoning	Lech Mazur benchmarks
51	GPT-6 Luna	OpenAI	68.7%	high reasoning	Lech Mazur benchmarks
52	Hy4 preview	Tencent	68.2%		Lech Mazur benchmarks
53	Claude Haiku 5.5	Anthropic	65.7%	high reasoning	Lech Mazur benchmarks
54	MiniMax-M3	MiniMax	65.1%		Lech Mazur benchmarks
55	GPT-5.4 mini	OpenAI	61.8%	xhigh reasoning	Lech Mazur benchmarks
56	Gemini 3.5 Flash Lite	Google	60.4%	high reasoning	Lech Mazur benchmarks
57	Qwen3.6 Plus	Alibaba (Qwen)	60.3%		Lech Mazur benchmarks
58	Qwen3.5 397B-A17B	Alibaba (Qwen)	58.9%		Lech Mazur benchmarks
59	Grok 4.3	xAI	55.2%		Lech Mazur benchmarks
60	Qwen3.8 27B	Alibaba (Qwen)	54.5%		Lech Mazur benchmarks
61	Claude Opus 4.5	Anthropic	52.5%		Lech Mazur benchmarks
62	Qwen3.5 122B-A10B	Alibaba (Qwen)	51.7%		Lech Mazur benchmarks
63	Qwen3.5 27B	Alibaba (Qwen)	47.9%		Lech Mazur benchmarks
64	Qwen3.7 Flash	Alibaba (Qwen)	43.8%		Lech Mazur benchmarks
65	Qwen3.6 35B-A3B	Alibaba (Qwen)	41.6%		Lech Mazur benchmarks
66	Hy3	Tencent	41.2%	high	Lech Mazur benchmarks
67	Step 3.7 Flash	StepFun	39.7%	high reasoning	Lech Mazur benchmarks
68	Claude Opus 4.7	Anthropic	39%	high reasoning	Lech Mazur benchmarks
69	Claude Sonnet 4.5	Anthropic	37.3%		Lech Mazur benchmarks
70	DeepSeek-V3.2-Exp	DeepSeek	36.7%		Lech Mazur benchmarks
71	MiMo-V2.5-Pro	Xiaomi	34.4%		Lech Mazur benchmarks
72	Qwen3 Max	Alibaba (Qwen)	30.1%	2026-01-23	Lech Mazur benchmarks
73	Seed 2.0 Pro	ByteDance Seed	28.4%		Lech Mazur benchmarks
74	Step 3.5 Flash	StepFun	28.4%		Lech Mazur benchmarks
75	Mistral Large 4	Mistral AI	27.4%	high	Lech Mazur benchmarks
76	MiMo-V2-Pro	Xiaomi	25.8%		Lech Mazur benchmarks
77	MiniMax-M2.7	MiniMax	24.7%		Lech Mazur benchmarks
78	ERNIE 5.1	Baidu	23.4%		Lech Mazur benchmarks
79	Longcat Flash Chat	Meituan	17.7%		Lech Mazur benchmarks
80	MiniMax-M2.5	MiniMax	16.8%		Lech Mazur benchmarks
81	Trinity Large Thinking	Arcee AI	16.5%		Lech Mazur benchmarks
82	Nemotron 3 Super	NVIDIA	15.4%		Lech Mazur benchmarks
83	MiniMax-M2	MiniMax	14.8%		Lech Mazur benchmarks
84	Claude Haiku 4.5	Anthropic	14.3%		Lech Mazur benchmarks
85	Mistral Medium 3.5	Mistral AI	12.9%	high	Lech Mazur benchmarks
86	MiniMax-M2.1	MiniMax	11.2%		Lech Mazur benchmarks
87	ERNIE 5.0 0110	Baidu	10.3%		Lech Mazur benchmarks
88	Gemini 3.1 Flash Lite	Google	8.2%		Lech Mazur benchmarks
89	Llama 4 Maverick	Meta	8%		Lech Mazur benchmarks
90	Mistral Large 3	Mistral AI	7.5%		Lech Mazur benchmarks
91	Mistral Medium 3.1	Mistral AI	6.5%		Lech Mazur benchmarks

Compare the leaders

Other reasoning benchmarks

Frequently asked questions

What does NYT Connections (extended) measure?

New York Times Connections puzzles made harder with extra distractor words: sort 16 or more words into groups of four by a hidden theme.

Which model has the highest NYT Connections (extended) score?

As of October 2026, GPT-6 Astra has the highest published NYT Connections (extended) score on Noometry at 98.1%, out of 91 models with results.

What is the best open-weight model on NYT Connections (extended)?

Kimi K3 has the highest NYT Connections (extended) accuracy among open-weight models at 93.6%, ranking 10 of 91 overall.