VoxelBench (text) · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	GPT-6 Astra	max	2734	–	–
2	GPT-6.1 Sol	max	2568	–	–
3	Opus 5.5	max	2549	–	–
4	GPT-6 Sol	max	2356	–	–
5	Opus 5	max	2269	–	–
6	Gemini 4	max	2230	–	–
7	GPT-5.6 Sol	max	2195	–	–
8	Fable 5	max	2097	–	–
9	Grok 4.6	xhigh	2046	–	–
10	GPT-5.5 Pro	xhigh	1970	–	–
11	GPT-5.5	xhigh	1954	–	–
12	Kimi K3	max	1926	–	–
13	GPT-5.6 Terra	max	1908	–	–
14	DeepSeek V4.1 Flash	max	1856	–	–
15	GPT-5.6 Luna	max	1848	–	–
16	Gemini 3.7 Flash	max	1810	–	–
17	Gemini 3.8 Flash	max	1809	–	–
18	GPT-6 Luna	max	1809	–	–
19	Qwen 3.8 Max	max	1783	–	–
20	Gemini 3.6 Flash	xhigh	1782	–	–
21	Haiku 5.5	max	1755	–	–
22	GLM 5.3	max	1752	–	–
23	GLM 5.3 Flash	max	1726	–	–
24	Grok 4.5	max	1718	–	–
25	Gemini 3.1 Pro	high	1701	–	–
26	Gemini 3.5 Flash	xhigh	1675	–	–
27	Muse Spark 1.3	max	1675	–	–
28	Opus 4.8	max	1600	–	–
29	DeepSeek V4 Pro	max	1599	–	–
30	GLM 5.2	max	1592	–	–
31	Gemini 2.5	–	1568	–	–
32	Qwen 3.7 Max	max	1552	–	–
33	Sonnet 5	xhigh	1547	–	–
34	Muse Spark 1.1	max	1525	–	–
35	Gemini 3	–	1454	–	–
36	GPT-5.2	xhigh	1441	–	–
37	GLM 5.1	max	1433	–	–
38	DeepSeek V4 Flash	max	1433	–	–
39	GPT-5.4	xhigh	1414	–	–
40	Opus 4.5	high	1401	–	–
41	Opus 4.6	max	1378	–	–
42	Gemini 3 Pro	high	1369	–	–
43	Opus 4.7	max	1345	–	–
44	MiniMax M3	max	1333	–	–
45	Qwen 3.6 Plus	max	1328	–	–
46	GPT-5	high	1314	–	–
47	Gemini 3 Flash	high	1302	–	–
48	GPT-5 Pro	high	1286	–	–
49	GPT-5 Codex	high	1269	–	–
50	Gemini 2.5 Pro	high	1267	–	–
51	Opus 4.1	high	1217	–	–
52	Sonnet 4.5	high	1202	–	–
53	GLM 5	high	1195	–	–
54	GPT-5.1	high	1185	–	–
55	Sonnet 5.5	max	1000	–	–