Best AI for research and checking facts: all 43 ranked models

Ordered by short factual questions. Score shown: how many short factual questions it answered correctly. Higher is better, and neighbouring models are often too close to separate: check the range before reading much into a one-place gap.

Show the other 28 ranked models

#	Model	Position	Score	Rating	Source
16	MeMuse Spark 1.1Meta	Ranked 16th of 43	57.8%standard error ±1.6 points	Strong	Source for Muse Spark 1.1 research a question score (epoch)
17	XaGrok 4.7xAI	Ranked 17th of 43	56.0%standard error ±1.6 points	Strong	Source for Grok 4.7 research a question score (epoch)
18	AnClaude Opus 4.8Anthropic	Ranked 18th of 43	53.0%standard error ±1.6 points	Strong	Source for Claude Opus 4.8 research a question score (epoch)
19	DeDeepSeek-V4-Pro (High, 20260813) Best open-weights option Open weightsDeepSeek	Ranked 19th of 43	52.9%standard error ±1.6 points	Strong	Source for DeepSeek-V4-Pro (High, 20260813) research a question score (epoch)
20	AlQwen3.6-Max-PreviewAlibaba	Ranked 20th of 43	52.0%standard error ±1.6 points	Strong	Source for Qwen3.6-Max-Preview research a question score (epoch)
21	AnClaude Opus 4.7Anthropic	Ranked 21st of 43	51.7%standard error ±1.6 points	Strong	Source for Claude Opus 4.7 research a question score (epoch)
22	MoKimi K3 Open weightsMoonshot	Ranked 22nd of 43	50.6%standard error ±1.6 points	Strong	Source for Kimi K3 research a question score (epoch)
23	XaGrok 4.6xAI	Ranked 23rd of 43	48.9%standard error ±1.6 points	Middle	Source for Grok 4.6 research a question score (epoch)
24	XaGrok 4.5xAI	Ranked 24th of 43	48.3%standard error ±1.6 points	Middle	Source for Grok 4.5 research a question score (epoch)
25	AnClaude Opus 4.6Anthropic	Ranked 25th of 43	47.0%standard error ±1.6 points	Middle	Source for Claude Opus 4.6 research a question score (epoch)
26	DeDeepSeek-V4-Pro Open weightsDeepSeek	Ranked 26th of 43	47.0%standard error ±1.6 points	Middle	Source for DeepSeek-V4-Pro research a question score (epoch)
27	AnClaude Sonnet 5.5Anthropic	Ranked 27th of 43	46.5%standard error ±1.6 points	Middle	Source for Claude Sonnet 5.5 research a question score (epoch)
28	AlQwen3.8-MaxAlibaba	Ranked 28th of 43	45.8%standard error ±1.6 points	Middle	Source for Qwen3.8-Max research a question score (epoch)
29	AlQwen3.6-PlusAlibaba	Ranked 29th of 43	44.1%standard error ±1.6 points	Middle	Source for Qwen3.6-Plus research a question score (epoch)
30	OpGPT-5.6 TerraOpenAI	Ranked 30th of 43	43.2%standard error ±1.6 points	Middle	Source for GPT-5.6 Terra research a question score (epoch)
31	OpGPT-6 LunaOpenAI	Ranked 31st of 43	41.4%standard error ±1.6 points	Middle	Source for GPT-6 Luna research a question score (epoch)
32	ZaGLM-5.3 Open weightsZ.ai (Zhipu AI)	Ranked 32nd of 43	41.0%standard error ±1.6 points	Middle	Source for GLM-5.3 research a question score (epoch)
32	OpGPT-5.6 LunaOpenAI	Ranked 32nd of 43	41.0%standard error ±1.6 points	Middle	Source for GPT-5.6 Luna research a question score (epoch)
34	MoKimi K2.6 Open weightsMoonshot	Ranked 34th of 43	34.9%standard error ±1.5 points	Behind the leaders	Source for Kimi K2.6 research a question score (epoch)
35	MoKimi K2.5 Open weightsMoonshot	Ranked 35th of 43	34.3%standard error ±1.5 points	Behind the leaders	Source for Kimi K2.5 research a question score (epoch)
36	ZaGLM-5.2 Open weightsZ.ai (Zhipu AI)	Ranked 36th of 43	34.2%standard error ±1.5 points	Behind the leaders	Source for GLM-5.2 research a question score (epoch)
37	ZaGLM-5.1 Open weightsZ.ai (Zhipu AI)	Ranked 37th of 43	34.0%standard error ±1.5 points	Behind the leaders	Source for GLM-5.1 research a question score (epoch)
38	AnClaude Sonnet 5Anthropic	Ranked 38th of 43	33.7%standard error ±1.5 points	Behind the leaders	Source for Claude Sonnet 5 research a question score (epoch)
39	AnClaude Sonnet 4.6Anthropic	Ranked 39th of 43	32.8%standard error ±1.5 points	Behind the leaders	Source for Claude Sonnet 4.6 research a question score (epoch)
40	ZaGLM-4.7 Open weightsZ.ai (Zhipu AI)	Ranked 40th of 43	32.2%standard error ±1.5 points	Behind the leaders	Source for GLM-4.7 research a question score (epoch)
41	AnClaude Haiku 5.5Anthropic	Ranked 41st of 43	23.8%standard error ±1.3 points	Behind the leaders	Source for Claude Haiku 5.5 research a question score (epoch)
42	MiMistral Large 4Mistral AI	Ranked 42nd of 43	20.0%standard error ±1.3 points	Behind the leaders	Source for Mistral Large 4 research a question score (epoch)
43	ThInkling-Small Open weightsThinking Machines	Ranked 43rd of 43	19.1%standard error ±1.2 points	Behind the leaders	Source for Inkling-Small research a question score (epoch)

No result for this test

Not enough data for short factual questions: amazon-nova-experimental-chat (26-02-10), Claude Opus 4.5 (20251101), DeepSeek-V4-Pro Preview, DeepSeek-V4.1-Flash-20260910, Dola Seed 2.0 Pro, ERNIE 5.0 (0110), ERNIE 5.1, Gemini 2.5 Pro, Gemini 3 Flash, Gemini 3 Pro, Gemini 3.1 Pro Preview, Gemini 3.5 Flash-Lite, Gemma 4 31B, GLM-5, GLM-5.3-Flash, GPT-5.1, GPT-5.4, GPT-5.5 Instant, Grok 4.20 Beta (Reasoning, 0309), Grok 4.20 Beta 1, Grok 4.20 Multi-Agent Beta (0309), Grok 4.3, MiMo-V2.5-Pro, MiMo-V2.6-Flash, MiMo-V2.6-Pro, Muse Spark, Muse Spark 1.3, Muse Wonder, NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4), Qwen3.5-397B-A17B, Qwen3.5-Flash, Qwen3.5-Max-Preview, Qwen3.7-Max-Preview and Qwen3.7-Plus. That is a gap in the evidence, not a low score.

Test: SimpleQA Verified. Source: epoch. Data last changed 29 Oct 2026. How we test.