AI Model Rankings

Measured success rate and latency for the models we serve, from real API calls made through ApexApi during health sweeps over the last 30 days. These are our own measured numbers, not vendor claims.

#	Model	Checks	Success	Avg latency	Last checked
1	Magistral Small 2509 bedrock	1	100%	1.8s	Sep 24
2	GPT-4.1 Nano openai	1	100%	1.8s	Sep 24
3	GPT-OSS 20B bedrock	1	100%	1.9s	Sep 24
4	Gemini 3.5 Flash Lite google-vertex	1	100%	2.0s	Sep 24
5	Claude Opus 4.7 anthropic	1	100%	2.1s	Sep 24
6	GPT-5.4 openai	1	100%	2.2s	Sep 24
7	Claude Opus 4.5 anthropic	1	100%	2.4s	Sep 24
8	Gemini 2.5 Flash Lite google	1	100%	2.4s	Sep 24
9	Ministral 3 8B 2512 mistral	1	100%	2.5s	Sep 24
10	Gemini 3.1 Flash Lite google	1	100%	2.5s	Sep 24
11	GPT-5.4 Nano openai	1	100%	2.5s	Sep 24
12	GPT-5.2 openai	1	100%	2.6s	Sep 24
13	Claude Opus 4.8 anthropic	1	100%	2.6s	Sep 24
14	GPT-4.1 Mini openai	1	100%	2.6s	Sep 24
15	Kimi K2.5 bedrock	1	100%	2.6s	Sep 24
16	GPT-5.4 Mini openai	1	100%	2.7s	Sep 24
17	GPT-4.1 openai	1	100%	2.7s	Sep 24
18	GPT-4o openai	1	100%	2.7s	Sep 24
19	Ministral 3 8B 2512 bedrock	1	100%	2.8s	Sep 24
20	Gemini 3.1 Flash Lite Preview google	1	100%	2.8s	Sep 24
21	Claude Haiku 4.5 anthropic	1	100%	2.9s	Sep 24
22	GPT-3.5 Turbo 16k openai	1	100%	2.9s	Sep 24
23	GLM 4.7 Flash bedrock	1	100%	3.0s	Sep 24
24	Codestral 2508 mistral	1	100%	3.0s	Sep 24
25	GPT-4o openai	1	100%	3.1s	Sep 24
26	Ministral 3 3B 2512 mistral	1	100%	3.1s	Sep 24
27	GPT-4o-mini openai	1	100%	3.1s	Sep 24
28	Gemini 3 Flash Preview google	1	100%	3.1s	Sep 24
29	Ministral 3 14B 2512 mistral	1	100%	3.2s	Sep 24
30	GPT-4o openai	1	100%	3.2s	Sep 24
31	Ministral 3 14B 2512 bedrock	1	100%	3.2s	Sep 24
32	Qwen3 VL Flash alibaba	1	100%	3.3s	Sep 24
33	DeepSeek V4 Pro deepseek	1	100%	3.3s	Sep 24
34	Claude Haiku 4.5 anthropic-vertex	1	100%	3.3s	Sep 24
35	Qwen3 VL Plus alibaba	1	100%	3.4s	Sep 24
36	GPT-OSS 120B bedrock	1	100%	3.4s	Sep 24
37	o3 openai	1	100%	3.4s	Sep 24
38	Qwen3 Coder Plus alibaba	1	100%	3.4s	Sep 24
39	Nano Banana google	1	100%	3.4s	Sep 24
40	Mistral Large 3 2512 bedrock	1	100%	3.4s	Sep 24
41	GPT-4o openai	1	100%	3.4s	Sep 24
42	Gemini 3.5 Flash google	1	100%	3.5s	Sep 24
43	o3 Mini openai	1	100%	3.5s	Sep 24
44	o4 Mini openai	1	100%	3.5s	Sep 24
45	Mistral Large 3 2512 mistral	1	100%	3.5s	Sep 24
46	GPT-4o-mini openai	1	100%	3.5s	Sep 24
47	GPT-5.6 Terra openai	1	100%	3.5s	Sep 24
48	Devstral 2 2512 bedrock	1	100%	3.6s	Sep 24
49	GPT-5.6 Luna openai	1	100%	3.7s	Sep 24
50	GPT-5.5 openai	1	100%	3.7s	Sep 24
51	GPT-3.5 Turbo openai	1	100%	3.8s	Sep 24
52	GPT-4 Turbo openai	1	100%	3.8s	Sep 24
53	Claude Opus 4.6 anthropic	1	100%	3.9s	Sep 24
54	Claude Opus 4.8 anthropic-vertex	1	100%	3.9s	Sep 24
55	o1 openai	1	100%	3.9s	Sep 24
56	Claude Opus 4.7 anthropic-vertex	1	100%	3.9s	Sep 24
57	Claude Opus 5 anthropic-vertex	1	100%	3.9s	Sep 24
58	Gemma 4 26B A4B google	1	100%	4.0s	Sep 24
59	Ministral 3 3B 2512 bedrock	1	100%	4.0s	Sep 24
60	GPT-5 Mini openai	1	100%	4.0s	Sep 24
61	GPT-5.1 openai	1	100%	4.0s	Sep 24
62	GLM 4.7 bedrock	1	100%	4.1s	Sep 24
63	GPT-4 openai	1	100%	4.2s	Sep 24
64	Devstral 2 2512 mistral	1	100%	4.2s	Sep 24
65	MiniMax M2.5 bedrock	1	100%	4.4s	Sep 24
66	Claude Sonnet 4.6 anthropic	1	100%	4.4s	Sep 24
67	Qwen3 Max alibaba	1	100%	4.5s	Sep 24
68	Claude Sonnet 4.6 anthropic-vertex	1	100%	4.6s	Sep 24
69	Gemma 4 31B google	1	100%	4.7s	Sep 24
70	Kimi K2 Thinking bedrock	1	100%	4.7s	Sep 24
71	Gemini 3.1 Pro Preview Custom Tools google	1	100%	4.7s	Sep 24
72	Claude Opus 4.6 anthropic-vertex	1	100%	4.8s	Sep 24
73	Gemini 3.7 Flash google-vertex	1	100%	5.0s	Sep 24
74	Qwen3.7 Max alibaba	1	100%	5.0s	Sep 24
75	Gemini 3.6 Flash google	1	100%	5.0s	Sep 24
76	Claude Opus 4.5 anthropic-vertex	1	100%	5.0s	Sep 24
77	Gemini 3.1 Pro Preview google	1	100%	5.1s	Sep 24
78	Qwen3.5 Flash alibaba	1	100%	5.3s	Sep 24
79	Grok 4.6 xai-vertex	1	100%	5.4s	Sep 24
80	GPT-5.6 Sol openai	1	100%	5.6s	Sep 24
81	Qwen3.6 Flash alibaba	1	100%	5.7s	Sep 24
82	Gemini 2.5 Flash google	1	100%	6.0s	Sep 24
83	GPT-5 Nano openai	1	100%	6.0s	Sep 24
84	Qwen3.7 Plus alibaba	1	100%	6.7s	Sep 24
85	GLM 5 bedrock	1	100%	7.4s	Sep 24
86	Gemini 2.5 Pro google	1	100%	8.4s	Sep 24
87	Qwen3.5 Plus alibaba	1	100%	15.4s	Sep 24
88	GPT-5 openai	1	100%	21.5s	Sep 24

Latency is the average of successful calls and depends on prompt size, region, and upstream load. Success rate reflects our smoke tests, not a guarantee. Call any of these through one API and key at the model catalog.