livebench coding typescript 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 Fable 5.1 Anthropic	60%	60.0	100% confidence 100 percent, Full
2	DeepSeek V4.1 Flash DeepSeek	60%	60.0	69% confidence 69 percent, Medium
3	Muse Spark 1.3 Meta	60%	60.0	93% confidence 93 percent, High
4	Qwen3.8 Max Alibaba / Qwen	56.7%	56.7	85% confidence 85 percent, High
5	DeepSeek V4 Flash Vision Exp DeepSeek	56.7%	56.7	21% confidence 21 percent, Low
6	Claude Fable 5 Anthropic	53.3%	53.3	100% confidence 100 percent, Full
7	Claude Opus 5.5 Anthropic	53.3%	53.3	100% confidence 100 percent, Full
8	Kimi K3 Moonshot AI	53.3%	53.3	100% confidence 100 percent, Full
9	Qwen3.8 27B Alibaba / Qwen	50%	50.0	69% confidence 69 percent, Medium
10	Claude Sonnet 5 Anthropic	50%	50.0	93% confidence 93 percent, High
11	Muse Spark 1.2 Meta	50%	50.0	53% confidence 53 percent, Medium
12	GPT-5.6 Sol OpenAI	50%	50.0	100% confidence 100 percent, Full
13	GLM-5.3 Z.ai	50%	50.0	93% confidence 93 percent, High
14	Qwen3.8 Flash Next Alibaba / Qwen	46.7%	46.7	21% confidence 21 percent, Low
15	Claude Sonnet 5.5 Anthropic	46.7%	46.7	93% confidence 93 percent, High
16	DeepSeek V4 Pro 0813 DeepSeek	46.7%	46.7	69% confidence 69 percent, Medium
17	Gemini 3.7 Flash Google	46.7%	46.7	100% confidence 100 percent, Full
18	GPT-5.6 Terra OpenAI	46.7%	46.7	100% confidence 100 percent, Full
19	GPT-6.1 Sol OpenAI	46.7%	46.7	100% confidence 100 percent, Full
20	GLM-5.3-Flash Z.ai	46.7%	46.7	100% confidence 100 percent, Full
21	Claude Haiku 5.5 Anthropic	43.3%	43.3	100% confidence 100 percent, Full
22	Claude Opus 4.7 Anthropic	43.3%	43.3	100% confidence 100 percent, Full
23	Claude Opus 5 Anthropic	43.3%	43.3	100% confidence 100 percent, Full
24	Muse Spark 1.1 Meta	43.3%	43.3	61% confidence 61 percent, Medium
25	Mistral Large 4 mistral	43.3%	43.3	32% confidence 32 percent, Low
26	GPT-5.4 OpenAI	43.3%	43.3	100% confidence 100 percent, Full
27	GPT-5.5 OpenAI	43.3%	43.3	100% confidence 100 percent, Full
28	GPT-6 Astra OpenAI	43.3%	43.3	100% confidence 100 percent, Full
29	Grok 4.6 xAI	43.3%	43.3	100% confidence 100 percent, Full
30	Grok 4.7 xAI	43.3%	43.3	100% confidence 100 percent, Full
31	Gemini 3.8 Flash Google	40%	40.0	100% confidence 100 percent, Full
32	GPT-6 Luna OpenAI	40%	40.0	100% confidence 100 percent, Full
33	GPT-6 Sol OpenAI	40%	40.0	100% confidence 100 percent, Full
34	GPT-5.2 OpenAI	36.7%	36.7	100% confidence 100 percent, Full
35	GPT-5.6 Luna OpenAI	36.7%	36.7	100% confidence 100 percent, Full
36	Grok 4.5 xAI	36.7%	36.7	100% confidence 100 percent, Full
37	GLM-5.2 Z.ai	36.7%	36.7	100% confidence 100 percent, Full
38	Claude Opus 4.6 Anthropic	33.3%	33.3	100% confidence 100 percent, Full
39	Claude Opus 4.8 Anthropic	33.3%	33.3	100% confidence 100 percent, Full
40	Gemini 3.5 Flash Google	33.3%	33.3	100% confidence 100 percent, Full
41	Grok Build 0.1 xAI	33.3%	33.3	16% confidence 16 percent, Low
42	Inkling thinkingmachines	30%	30.0	100% confidence 100 percent, Full
43	Qwen3.7 Max Alibaba / Qwen	26.7%	26.7	53% confidence 53 percent, Medium
44	DeepSeek V4 Flash 0731 DeepSeek	26.7%	26.7	69% confidence 69 percent, Medium
45	Gemini 3.5 Flash Lite Google	26.7%	26.7	100% confidence 100 percent, Full
46	Gemini 3.6 Flash Google	26.7%	26.7	100% confidence 100 percent, Full
47	Kimi K2.6 Moonshot AI	26.7%	26.7	85% confidence 85 percent, High
48	GPT-5.4 nano OpenAI	26.7%	26.7	100% confidence 100 percent, Full
49	Qwen3.6 27B Alibaba / Qwen	23.3%	23.3	53% confidence 53 percent, Medium
50	Claude Sonnet 4.6 Anthropic	23.3%	23.3	100% confidence 100 percent, Full
51	DeepSeek V4 Flash DeepSeek	23.3%	23.3	53% confidence 53 percent, Medium
52	DeepSeek V4 Pro DeepSeek	23.3%	23.3	85% confidence 85 percent, High
53	Gemini 3.1 Pro Preview Google	23.3%	23.3	100% confidence 100 percent, Full
54	MiniMax-M3 minimax	23.3%	23.3	100% confidence 100 percent, Full
55	Kimi K2.7 Code Moonshot AI	23.3%	23.3	48% confidence 48 percent, Low
56	Qwen3.6 Plus Alibaba / Qwen	20%	20.0	80% confidence 80 percent, High
57	Claude Opus 4.5 Anthropic	20%	20.0	100% confidence 100 percent, Full
58	GPT-5.2 Codex OpenAI	20%	20.0	43% confidence 43 percent, Low
59	GPT-5.4 mini OpenAI	20%	20.0	100% confidence 100 percent, Full
60	Nemotron 3 Ultra 550B A55B nvidia	16.7%	16.7	100% confidence 100 percent, Full
61	Grok 4.3 xAI	13.3%	13.3	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.