Terminal-Bench 4.0 (Vals) · Open Analysis

#	Model	Effort	Score	Cost / task	Harness
1	Opus 5.5	max	65.2%	$13.20	–
2	Sonnet 5.5	max	64.1%	$16.51	–
3	GPT-6 Astra	max	59.6%	$9.58	–
4	Fable 5.1	max	58.1%	$17.18	–
5	Wonder Full	max	58.1%	$10.53	–
6	Gemini 4	high	57.6%	$17.64	–
7	GPT-6.1 Sol	max	55.1%	$1.72	–
8	Opus 5	max	53.5%	$18.60	–
9	Wonder	max	49.5%	$14.66	–
10	GPT-6 Sol	max	44.4%	$5.79	–
11	Fable 5	max	41.4%	$30.34	–
12	GLM 5.3	max	38.9%	$9.37	–
13	GPT-5.6 Sol	max	37.9%	$7.98	–
14	GLM 5.3 Flash	max	37.4%	$1.22	–
15	Haiku 5.5	max	36.4%	$3.12	–
16	Qwen 3.8 Max	max	34.3%	$10.63	–
17	Step 5 Preview	–	31.8%	$3.11	–
18	MiMo V2.6 Pro	high	31.3%	$0.498	–
19	Grok 4.7	xhigh	28.8%	$18.01	–
20	Muse Spark 1.3	max	24.7%	$6.65	–
21	MiMo V2.6 Flash	high	24.2%	$0.216	–
22	Mistral Large 4	high	23.2%	$7.90	–
23	Opus 4.8	max	23.2%	$17.14	–
24	GPT-5.6 Terra	max	22.7%	$5.60	–
25	Ember-1	–	20.7%	$19.48	–
26	DeepSeek V4.1 Flash	high	19.7%	$0.499	–
27	Gemini 3.8 Flash	high	19.7%	$8.65	–
28	DeepSeek V4 Flash 0731	high	18.7%	$0.766	–
29	Grok 4.6	high	17.2%	$5.06	–
30	Kimi K3	max	17.2%	$12.48	–
31	DeepSeek V4 Pro 0813	max	14.1%	$3.31	–
32	GPT-6 Luna	max	13.6%	$0.355	–
33	Gemini 3.7 Flash	high	12.1%	$8.71	–
34	GPT-5.6 Luna	max	11.6%	$0.732	–
35	DeepSeek V4	max	11.1%	$3.45	–
36	Sonnet 5	max	10.6%	$26.05	–
37	Grok 4.5	high	8.6%	$5.52	–
38	Hy4	–	8.1%	$2.10	–
39	Gemini 3.5 Flash	high	6.1%	$5.98	–
40	Muse Spark 1.2	xhigh	6.1%	$4.14	–
41	Gemini 3.1 Pro	high	2.5%	$4.11	–
42	GPT-5.4 Mini	xhigh	2.5%	$1.44	–
43	MiniMax M3	max	1.0%	$7.23	–
44	Inkling	max	0.5%	$2.50	–
45	MiMo V2.5 Pro	xhigh	0.5%	$0.233	–
46	Inkling Small	xhigh	0.0%	$0.322	–
47	Mercury 2.5	high	0.0%	$0.368	–