Code Migration

Can language models reimplement real-world programs in another language? Vals AI's overall score for Code Migration.

58 models · 59 results · updated 2026-10-18 · vals.ai ↗

More from this benchmark

1 published measures

Time per task79 results · lower is better

Highlight models

Click a point on the chart to highlight that exact effort; click it again to remove it.

43 more in settings

Chart settings

58 of 58 selected

59 rows

#	Model	Effort	Code Migration	Subscription $ / task	API $ / task	Time
1	Sonnet 5.5	max	69.8%	$1.15	$76	3.3 h
2	Gemini 4 Argon	high	68.2%	$21	$58	2.2 h
3	GPT-6 Astra	max	67.7%	$3.80	$44	59 min
4	Opus 5.5	max	66.6%	$1.71	$113	3.3 h
5	GPT-6.1 Sol	max	65.1%	$0.76	$6.51	1.6 h
6	Opus 5	max	57.5%	$1.05	$61	2.9 h
7	GPT-6 Sol	max	57.2%	$0.97	$16	1.4 h
8	Fable 5	max	55.1%	$13	$112	1.8 h
9	Fable 5.1	max	54.6%	$11	$71	4.3 h
10	GPT-5.6 Sol	max	52.9%	$1.36	$25	59 min
11	GPT-5.6 Terra	max	47.8%	$0.45	$8.13	49 min
12	Muse Spark 1.3	max	47.4%	$7.11	$14	1.4 h
13	Opus 4.8	max	47.3%	$0.53	$31	1.2 h
14	DeepSeek V4.1 Flash	high	45.6%	$0.059	$0.94	1.8 h
15	GPT-5.5	xhigh	45.2%	$0.36	$6.44	32 min
16	Grok 4.7	xhigh	44.8%	$3.39	$37	1.2 h
17	Grok 4.6	high	44.6%	$1.50	$15	1.5 h
18	Haiku 5.5	max	44.6%	$0.13	$6.23	1.4 h
19	GPT-5.6 Luna	max	44.5%	$0.18	$1.88	59 min
20	Sonnet 5	max	44.4%	$0.61	$35	2.0 h
21	GLM 5.3	max	44.2%	$2.59	$25	3.8 h
22	DeepSeek V4.1 Flash 1017	max	44.2%	free	$1.86	2.2 h
23	Opus 4.7	max	43.9%	$36	$36	1.2 h
24	MiMo V2.6 Pro	high	43.0%	$0.58	$0.69	2.8 h
25	GPT-6 Luna	max	42.6%	$0.089	$0.60	50 min
26	DeepSeek V4 Pro 0813	max	41.5%	$19	$19	3.5 h
27	MiMo V2.6 Flash	high	40.9%	free	$0.49	3.2 h
28	Sonnet 4.6	max	39.9%	$39	$39	1.7 h
29	DeepSeek V4 Flash 0731	high	38.6%	$3.61	$3.61	2.1 h
30	GLM 5.2	max	37.9%	$1.10	$13	1.9 h