livebench coding code completion 2026_06_25 Benchmark: Scores and Sources

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

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

Row	Model	Result	Normalized (0–100)	Confidence
1	Claude Opus 5.5 Anthropic	87.0%	87.0	100% confidence 100 percent, Full
2	GPT-5.2 Codex OpenAI	87.0%	87.0	43% confidence 43 percent, Low
3	GPT-5.6 Luna OpenAI	87.0%	87.0	100% confidence 100 percent, Full
4	Claude Opus 4.8 Anthropic	84.8%	84.8	100% confidence 100 percent, Full
5	Claude Sonnet 5.5 Anthropic	84.8%	84.8	93% confidence 93 percent, High
6	GPT-5.6 Sol OpenAI	84.8%	84.8	100% confidence 100 percent, Full
7	Claude Fable 5.1 Anthropic	82.6%	82.6	100% confidence 100 percent, Full
8	Claude Opus 5 Anthropic	82.6%	82.6	100% confidence 100 percent, Full
9	DeepSeek V4.1 Flash DeepSeek	82.6%	82.6	69% confidence 69 percent, Medium
10	Kimi K3 Moonshot AI	82.6%	82.6	100% confidence 100 percent, Full
11	GPT-5.5 OpenAI	82.6%	82.6	100% confidence 100 percent, Full
12	GPT-6.1 Sol OpenAI	82.6%	82.6	100% confidence 100 percent, Full
13	Claude Fable 5 Anthropic	80.4%	80.4	100% confidence 100 percent, Full
14	Claude Opus 4.5 Anthropic	80.4%	80.4	100% confidence 100 percent, Full
15	Muse Spark 1.2 Meta	80.4%	80.4	53% confidence 53 percent, Medium
16	Muse Spark 1.3 Meta	80.4%	80.4	93% confidence 93 percent, High
17	GPT-5.4 OpenAI	80.4%	80.4	100% confidence 100 percent, Full
18	GPT-5.6 Terra OpenAI	80.4%	80.4	100% confidence 100 percent, Full
19	GPT-6 Astra OpenAI	80.4%	80.4	100% confidence 100 percent, Full
20	GPT-6 Luna OpenAI	80.4%	80.4	100% confidence 100 percent, Full
21	GPT-6 Sol OpenAI	80.4%	80.4	100% confidence 100 percent, Full
22	GLM-5.2 Z.ai	80.4%	80.4	100% confidence 100 percent, Full
23	GLM-5.3 Z.ai	80.4%	80.4	93% confidence 93 percent, High
24	GLM-5.3-Flash Z.ai	80.4%	80.4	100% confidence 100 percent, Full
25	Claude Haiku 5.5 Anthropic	78.3%	78.3	100% confidence 100 percent, Full
26	Claude Opus 4.7 Anthropic	78.3%	78.3	100% confidence 100 percent, Full
27	Claude Sonnet 4.6 Anthropic	78.3%	78.3	100% confidence 100 percent, Full
28	Claude Sonnet 5 Anthropic	78.3%	78.3	93% confidence 93 percent, High
29	DeepSeek V4 Pro 0813 DeepSeek	78.3%	78.3	69% confidence 69 percent, Medium
30	Gemini 3.1 Pro Preview Google	78.3%	78.3	100% confidence 100 percent, Full
31	Gemini 3.6 Flash Google	78.3%	78.3	100% confidence 100 percent, Full
32	Muse Spark 1.1 Meta	78.3%	78.3	61% confidence 61 percent, Medium
33	Mistral Large 4 mistral	78.3%	78.3	32% confidence 32 percent, Low
34	Kimi K2.6 Moonshot AI	78.3%	78.3	85% confidence 85 percent, High
35	Grok 4.7 xAI	78.3%	78.3	100% confidence 100 percent, Full
36	Qwen3.6 Plus Alibaba / Qwen	76.1%	76.1	80% confidence 80 percent, High
37	Qwen3.8 Flash Next Alibaba / Qwen	76.1%	76.1	21% confidence 21 percent, Low
38	Claude Opus 4.6 Anthropic	76.1%	76.1	100% confidence 100 percent, Full
39	Gemini 3.5 Flash Google	76.1%	76.1	100% confidence 100 percent, Full
40	Gemini 3.5 Flash Lite Google	76.1%	76.1	100% confidence 100 percent, Full
41	Gemini 3.7 Flash Google	76.1%	76.1	100% confidence 100 percent, Full
42	Kimi K2.7 Code Moonshot AI	76.1%	76.1	48% confidence 48 percent, Low
43	GPT-5.2 OpenAI	76.1%	76.1	100% confidence 100 percent, Full
44	Grok 4.6 xAI	76.1%	76.1	100% confidence 100 percent, Full
45	GPT-5.4 mini OpenAI	75.6%	75.6	100% confidence 100 percent, Full
46	Qwen3.8 27B Alibaba / Qwen	73.9%	73.9	69% confidence 69 percent, Medium
47	Qwen3.8 Max Alibaba / Qwen	73.9%	73.9	85% confidence 85 percent, High
48	DeepSeek V4 Flash 0731 DeepSeek	73.9%	73.9	69% confidence 69 percent, Medium
49	Qwen3.6 27B Alibaba / Qwen	71.7%	71.7	53% confidence 53 percent, Medium
50	Gemini 3.8 Flash Google	71.7%	71.7	100% confidence 100 percent, Full
51	Qwen3.7 Max Alibaba / Qwen	69.6%	69.6	53% confidence 53 percent, Medium
52	DeepSeek V4 Pro DeepSeek	69.6%	69.6	85% confidence 85 percent, High
53	Nemotron 3 Ultra 550B A55B nvidia	69.6%	69.6	100% confidence 100 percent, Full
54	Grok 4.5 xAI	69.6%	69.6	100% confidence 100 percent, Full
55	DeepSeek V4 Flash Vision Exp DeepSeek	67.4%	67.4	21% confidence 21 percent, Low
56	MiniMax-M3 minimax	67.4%	67.4	100% confidence 100 percent, Full
57	GPT-5.4 nano OpenAI	67.4%	67.4	100% confidence 100 percent, Full
58	Inkling thinkingmachines	67.4%	67.4	100% confidence 100 percent, Full
59	Grok Build 0.1 xAI	67.4%	67.4	16% confidence 16 percent, Low
60	DeepSeek V4 Flash DeepSeek	65.2%	65.2	53% confidence 53 percent, Medium
61	Grok 4.3 xAI	65.2%	65.2	80% confidence 80 percent, High

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 coding pillar of the SI Score.