livebench reasoning logic with navigation 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	Muse Spark 1.2 Meta	88%	88.0	53% confidence 53 percent, Medium
2	GPT-6 Astra OpenAI	88%	88.0	100% confidence 100 percent, Full
3	Claude Fable 5.1 Anthropic	86%	86.0	100% confidence 100 percent, Full
4	Claude Opus 5 Anthropic	86%	86.0	100% confidence 100 percent, Full
5	Claude Sonnet 5 Anthropic	86%	86.0	93% confidence 93 percent, High
6	Qwen3.7 Max Alibaba / Qwen	84%	84.0	53% confidence 53 percent, Medium
7	Gemini 3.8 Flash Google	82%	82.0	100% confidence 100 percent, Full
8	GPT-5.6 Sol OpenAI	82%	82.0	100% confidence 100 percent, Full
9	GPT-6.1 Sol OpenAI	82%	82.0	100% confidence 100 percent, Full
10	Grok 4.6 xAI	82%	82.0	100% confidence 100 percent, Full
11	Claude Opus 4.6 Anthropic	80%	80.0	100% confidence 100 percent, Full
12	Claude Opus 5.5 Anthropic	80%	80.0	100% confidence 100 percent, Full
13	Claude Sonnet 5.5 Anthropic	80%	80.0	93% confidence 93 percent, High
14	Kimi K3 Moonshot AI	80%	80.0	100% confidence 100 percent, Full
15	GPT-5.6 Luna OpenAI	80%	80.0	100% confidence 100 percent, Full
16	GLM-5.2 Z.ai	80%	80.0	100% confidence 100 percent, Full
17	Claude Fable 5 Anthropic	78%	78.0	100% confidence 100 percent, Full
18	Claude Opus 4.8 Anthropic	78%	78.0	100% confidence 100 percent, Full
19	Muse Spark 1.1 Meta	78%	78.0	61% confidence 61 percent, Medium
20	Gemini 3.6 Flash Google	76%	76.0	100% confidence 100 percent, Full
21	Muse Spark 1.3 Meta	76%	76.0	93% confidence 93 percent, High
22	GPT-5.5 OpenAI	76%	76.0	100% confidence 100 percent, Full
23	GPT-5.6 Terra OpenAI	76%	76.0	100% confidence 100 percent, Full
24	Qwen3.8 Max Alibaba / Qwen	74%	74.0	85% confidence 85 percent, High
25	Gemini 3.5 Flash Google	74%	74.0	100% confidence 100 percent, Full
26	Gemini 3.7 Flash Google	74%	74.0	100% confidence 100 percent, Full
27	Kimi K2.7 Code Moonshot AI	74%	74.0	48% confidence 48 percent, Low
28	GPT-5.2 OpenAI	74%	74.0	100% confidence 100 percent, Full
29	GPT-6 Sol OpenAI	74%	74.0	100% confidence 100 percent, Full
30	Gemini 3.1 Pro Preview Google	72%	72.0	100% confidence 100 percent, Full
31	Grok 4.5 xAI	72%	72.0	100% confidence 100 percent, Full
32	Qwen3.6 Plus Alibaba / Qwen	70%	70.0	80% confidence 80 percent, High
33	Qwen3.8 Flash Next Alibaba / Qwen	70%	70.0	21% confidence 21 percent, Low
34	Claude Opus 4.7 Anthropic	70%	70.0	100% confidence 100 percent, Full
35	Claude Sonnet 4.6 Anthropic	70%	70.0	100% confidence 100 percent, Full
36	DeepSeek V4 Flash 0731 DeepSeek	70%	70.0	69% confidence 69 percent, Medium
37	Kimi K2.6 Moonshot AI	70%	70.0	85% confidence 85 percent, High
38	Grok Build 0.1 xAI	70%	70.0	16% confidence 16 percent, Low
39	GLM-5.3 Z.ai	70%	70.0	93% confidence 93 percent, High
40	Qwen3.8 27B Alibaba / Qwen	68%	68.0	69% confidence 69 percent, Medium
41	Claude Opus 4.5 Anthropic	68%	68.0	100% confidence 100 percent, Full
42	DeepSeek V4 Flash Vision Exp DeepSeek	68%	68.0	21% confidence 21 percent, Low
43	DeepSeek V4.1 Flash DeepSeek	68%	68.0	69% confidence 69 percent, Medium
44	Gemini 3.5 Flash Lite Google	68%	68.0	100% confidence 100 percent, Full
45	Mistral Large 4 mistral	68%	68.0	32% confidence 32 percent, Low
46	Nemotron 3 Ultra 550B A55B nvidia	68%	68.0	100% confidence 100 percent, Full
47	GPT-5.2 Codex OpenAI	68%	68.0	43% confidence 43 percent, Low
48	GPT-5.4 OpenAI	68%	68.0	100% confidence 100 percent, Full
49	Claude Haiku 5.5 Anthropic	66%	66.0	100% confidence 100 percent, Full
50	DeepSeek V4 Pro 0813 DeepSeek	66%	66.0	69% confidence 69 percent, Medium
51	MiniMax-M3 minimax	66%	66.0	100% confidence 100 percent, Full
52	Inkling thinkingmachines	66%	66.0	100% confidence 100 percent, Full
53	Grok 4.3 xAI	66%	66.0	80% confidence 80 percent, High
54	Grok 4.7 xAI	66%	66.0	100% confidence 100 percent, Full
55	DeepSeek V4 Flash DeepSeek	64%	64.0	53% confidence 53 percent, Medium
56	DeepSeek V4 Pro DeepSeek	64%	64.0	85% confidence 85 percent, High
57	GPT-5.4 nano OpenAI	64%	64.0	100% confidence 100 percent, Full
58	Qwen3.6 27B Alibaba / Qwen	62%	62.0	53% confidence 53 percent, Medium
59	GPT-6 Luna OpenAI	62%	62.0	100% confidence 100 percent, Full
60	GLM-5.3-Flash Z.ai	62%	62.0	100% confidence 100 percent, Full
61	GPT-5.4 mini OpenAI	60%	60.0	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.