livebench reasoning consecutive events 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	Claude Sonnet 4.6 Anthropic	92.7%	92.7	100% confidence 100 percent, Full
2	Claude Fable 5 Anthropic	91.4%	91.4	100% confidence 100 percent, Full
3	GPT-6.1 Sol OpenAI	91.3%	91.3	100% confidence 100 percent, Full
4	Claude Sonnet 5.5 Anthropic	90.9%	90.9	93% confidence 93 percent, High
5	Claude Fable 5.1 Anthropic	90.6%	90.6	100% confidence 100 percent, Full
6	GPT-6 Astra OpenAI	90.6%	90.6	100% confidence 100 percent, Full
7	Muse Spark 1.3 Meta	90.5%	90.5	93% confidence 93 percent, High
8	Claude Opus 5.5 Anthropic	90.4%	90.4	100% confidence 100 percent, Full
9	DeepSeek V4 Pro 0813 DeepSeek	90.2%	90.2	69% confidence 69 percent, Medium
10	GPT-5.6 Sol OpenAI	90.2%	90.2	100% confidence 100 percent, Full
11	DeepSeek V4 Flash Vision Exp DeepSeek	90.0%	90.0	21% confidence 21 percent, Low
12	Kimi K3 Moonshot AI	89.8%	89.8	100% confidence 100 percent, Full
13	GPT-6 Sol OpenAI	89.5%	89.5	100% confidence 100 percent, Full
14	Claude Opus 4.7 Anthropic	89.5%	89.5	100% confidence 100 percent, Full
15	DeepSeek V4 Flash 0731 DeepSeek	89.4%	89.4	69% confidence 69 percent, Medium
16	GPT-5.2 Codex OpenAI	89.0%	89.0	43% confidence 43 percent, Low
17	GPT-5.2 OpenAI	88.9%	88.9	100% confidence 100 percent, Full
18	GPT-5.5 OpenAI	88.8%	88.8	100% confidence 100 percent, Full
19	DeepSeek V4.1 Flash DeepSeek	88.7%	88.7	69% confidence 69 percent, Medium
20	GPT-5.6 Terra OpenAI	87.7%	87.7	100% confidence 100 percent, Full
21	Qwen3.8 Max Alibaba / Qwen	87.1%	87.1	85% confidence 85 percent, High
22	GPT-5.6 Luna OpenAI	86.5%	86.5	100% confidence 100 percent, Full
23	GPT-5.4 OpenAI	86.2%	86.2	100% confidence 100 percent, Full
24	Grok 4.7 xAI	85.7%	85.7	100% confidence 100 percent, Full
25	Gemini 3.1 Pro Preview Google	85.2%	85.2	100% confidence 100 percent, Full
26	Qwen3.8 27B Alibaba / Qwen	84.9%	84.9	69% confidence 69 percent, Medium
27	MiniMax-M3 minimax	84.6%	84.6	100% confidence 100 percent, Full
28	GLM-5.3-Flash Z.ai	84.3%	84.3	100% confidence 100 percent, Full
29	Muse Spark 1.2 Meta	82.5%	82.5	53% confidence 53 percent, Medium
30	Mistral Large 4 mistral	80.5%	80.5	32% confidence 32 percent, Low
31	Claude Opus 4.5 Anthropic	79.4%	79.4	100% confidence 100 percent, Full
32	GLM-5.2 Z.ai	79.3%	79.3	100% confidence 100 percent, Full
33	Grok 4.6 xAI	78.1%	78.1	100% confidence 100 percent, Full
34	Claude Opus 5 Anthropic	77.6%	77.6	100% confidence 100 percent, Full
35	Qwen3.8 Flash Next Alibaba / Qwen	75.6%	75.6	21% confidence 21 percent, Low
36	Grok 4.5 xAI	75.5%	75.5	100% confidence 100 percent, Full
37	DeepSeek V4 Pro DeepSeek	75.4%	75.4	85% confidence 85 percent, High
38	Claude Sonnet 5 Anthropic	74.7%	74.7	93% confidence 93 percent, High
39	GLM-5.3 Z.ai	72.3%	72.3	93% confidence 93 percent, High
40	Qwen3.7 Max Alibaba / Qwen	71.8%	71.8	53% confidence 53 percent, Medium
41	Muse Spark 1.1 Meta	71.1%	71.1	61% confidence 61 percent, Medium
42	GPT-6 Luna OpenAI	70.5%	70.5	100% confidence 100 percent, Full
43	Qwen3.6 27B Alibaba / Qwen	70.4%	70.4	53% confidence 53 percent, Medium
44	Inkling thinkingmachines	69.9%	69.9	100% confidence 100 percent, Full
45	Grok Build 0.1 xAI	69.7%	69.7	16% confidence 16 percent, Low
46	Qwen3.6 Plus Alibaba / Qwen	67.0%	67.0	80% confidence 80 percent, High
47	Claude Opus 4.6 Anthropic	63.5%	63.5	100% confidence 100 percent, Full
48	GPT-5.4 mini OpenAI	63.0%	63.0	100% confidence 100 percent, Full
49	Gemini 3.7 Flash Google	60.4%	60.4	100% confidence 100 percent, Full
50	DeepSeek V4 Flash DeepSeek	59.9%	59.9	53% confidence 53 percent, Medium
51	GPT-5.4 nano OpenAI	54.3%	54.3	100% confidence 100 percent, Full
52	Claude Opus 4.8 Anthropic	52.9%	52.9	100% confidence 100 percent, Full
53	Kimi K2.6 Moonshot AI	51.3%	51.3	85% confidence 85 percent, High
54	Kimi K2.7 Code Moonshot AI	48.3%	48.3	48% confidence 48 percent, Low
55	Gemini 3.5 Flash Google	48.0%	48.0	100% confidence 100 percent, Full
56	Gemini 3.6 Flash Google	43.8%	43.8	100% confidence 100 percent, Full
57	Grok 4.3 xAI	28.0%	28.0	80% confidence 80 percent, High
58	Gemini 3.5 Flash Lite Google	21.1%	21.1	100% confidence 100 percent, Full
59	Gemini 3.8 Flash Google	20.6%	20.6	100% confidence 100 percent, Full
60	Nemotron 3 Ultra 550B A55B nvidia	18.8%	18.8	100% confidence 100 percent, Full
61	Claude Haiku 5.5 Anthropic	10.6%	10.6	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.