livebench reasoning zebra puzzle 2026_06_25 Benchmark: Scores and Sources

Published result; benchmark version and evaluation conditions remain in the id and result note.

pillar: reasoning · weight 1 within pillar · unit: % (higher is better) · official board ↗

Row	Model	Result	Normalized (0–100)	Confidence
1	Qwen3.8 Max Alibaba / Qwen	100%	100.0	85% confidence 85 percent, High
2	Claude Fable 5 Anthropic	100%	100.0	100% confidence 100 percent, Full
3	Claude Fable 5.1 Anthropic	100%	100.0	100% confidence 100 percent, Full
4	Claude Opus 4.8 Anthropic	100%	100.0	100% confidence 100 percent, Full
5	Claude Opus 5 Anthropic	100%	100.0	100% confidence 100 percent, Full
6	Claude Opus 5.5 Anthropic	100%	100.0	100% confidence 100 percent, Full
7	Claude Sonnet 5.5 Anthropic	100%	100.0	93% confidence 93 percent, High
8	DeepSeek V4 Flash 0731 DeepSeek	100%	100.0	69% confidence 69 percent, Medium
9	DeepSeek V4.1 Flash DeepSeek	100%	100.0	69% confidence 69 percent, Medium
10	Kimi K3 Moonshot AI	100%	100.0	100% confidence 100 percent, Full
11	GPT-5.5 OpenAI	100%	100.0	100% confidence 100 percent, Full
12	GPT-5.6 Sol OpenAI	100%	100.0	100% confidence 100 percent, Full
13	GPT-5.6 Terra OpenAI	100%	100.0	100% confidence 100 percent, Full
14	GPT-6 Astra OpenAI	100%	100.0	100% confidence 100 percent, Full
15	GPT-6 Sol OpenAI	100%	100.0	100% confidence 100 percent, Full
16	GPT-6.1 Sol OpenAI	100%	100.0	100% confidence 100 percent, Full
17	Qwen3.8 Flash Next Alibaba / Qwen	98.8%	98.8	21% confidence 21 percent, Low
18	DeepSeek V4 Flash Vision Exp DeepSeek	98.8%	98.8	21% confidence 21 percent, Low
19	Gemini 3.8 Flash Google	98.3%	98.3	100% confidence 100 percent, Full
20	Claude Opus 4.7 Anthropic	98%	98.0	100% confidence 100 percent, Full
21	Muse Spark 1.3 Meta	98%	98.0	93% confidence 93 percent, High
22	GPT-5.4 OpenAI	98%	98.0	100% confidence 100 percent, Full
23	GPT-5.4 nano OpenAI	97.6%	97.6	100% confidence 100 percent, Full
24	DeepSeek V4 Pro 0813 DeepSeek	96.8%	96.8	69% confidence 69 percent, Medium
25	Qwen3.8 27B Alibaba / Qwen	96.5%	96.5	69% confidence 69 percent, Medium
26	Claude Sonnet 4.6 Anthropic	96%	96.0	100% confidence 100 percent, Full
27	Muse Spark 1.1 Meta	96%	96.0	61% confidence 61 percent, Medium
28	Kimi K2.7 Code Moonshot AI	96%	96.0	48% confidence 48 percent, Low
29	GPT-6 Luna OpenAI	96%	96.0	100% confidence 100 percent, Full
30	Grok 4.6 xAI	95.5%	95.5	100% confidence 100 percent, Full
31	Muse Spark 1.2 Meta	95.3%	95.3	53% confidence 53 percent, Medium
32	Gemini 3.7 Flash Google	94.5%	94.5	100% confidence 100 percent, Full
33	Claude Haiku 5.5 Anthropic	94%	94.0	100% confidence 100 percent, Full
34	Claude Opus 4.6 Anthropic	94%	94.0	100% confidence 100 percent, Full
35	Grok 4.5 xAI	94%	94.0	100% confidence 100 percent, Full
36	GPT-5.6 Luna OpenAI	93.5%	93.5	100% confidence 100 percent, Full
37	Mistral Large 4 mistral	93%	93.0	32% confidence 32 percent, Low
38	DeepSeek V4 Pro DeepSeek	92%	92.0	85% confidence 85 percent, High
39	GLM-5.3 Z.ai	88.8%	88.8	93% confidence 93 percent, High
40	Claude Sonnet 5 Anthropic	88%	88.0	93% confidence 93 percent, High
41	Inkling thinkingmachines	88%	88.0	100% confidence 100 percent, Full
42	Grok 4.7 xAI	86%	86.0	100% confidence 100 percent, Full
43	Gemini 3.6 Flash Google	85.8%	85.8	100% confidence 100 percent, Full
44	Gemini 3.1 Pro Preview Google	85.3%	85.3	100% confidence 100 percent, Full
45	GPT-5.2 OpenAI	84%	84.0	100% confidence 100 percent, Full
46	GLM-5.3-Flash Z.ai	79.5%	79.5	100% confidence 100 percent, Full
47	Kimi K2.6 Moonshot AI	78.5%	78.5	85% confidence 85 percent, High
48	Claude Opus 4.5 Anthropic	77.5%	77.5	100% confidence 100 percent, Full
49	Gemini 3.5 Flash Google	77.3%	77.3	100% confidence 100 percent, Full
50	Qwen3.7 Max Alibaba / Qwen	74.5%	74.5	53% confidence 53 percent, Medium
51	Nemotron 3 Ultra 550B A55B nvidia	71.3%	71.3	100% confidence 100 percent, Full
52	GPT-5.2 Codex OpenAI	70%	70.0	43% confidence 43 percent, Low
53	Grok Build 0.1 xAI	68.3%	68.3	16% confidence 16 percent, Low
54	Qwen3.6 Plus Alibaba / Qwen	68%	68.0	80% confidence 80 percent, High
55	GLM-5.2 Z.ai	65.5%	65.5	100% confidence 100 percent, Full
56	MiniMax-M3 minimax	61%	61.0	100% confidence 100 percent, Full
57	GPT-5.4 mini OpenAI	60.4%	60.4	100% confidence 100 percent, Full
58	Grok 4.3 xAI	57.8%	57.8	80% confidence 80 percent, High
59	Qwen3.6 27B Alibaba / Qwen	53.8%	53.8	53% confidence 53 percent, Medium
60	DeepSeek V4 Flash DeepSeek	49.3%	49.3	53% confidence 53 percent, Medium
61	Gemini 3.5 Flash Lite Google	32.8%	32.8	100% confidence 100 percent, Full

Results are as published by the source behind each value (hover or tap the number). Benchmark scores are shown individually for every source/variant (a model may have multiple rows), and vary by version, harness and date; the normalized column uses the method’s fixed 0–100 scales and feeds the reasoning pillar of the SI Score.