模型排行

排名	变化	模型	得分	置信区间
1	—	Sonic 3.6闭源Cartesia	1,273	±17
2	新	Gemini 3.8 Flash TTS闭源Google	1,260	±17
3	—	Qwen-Audio-3.0-TTS-Plus闭源Alibaba	1,259	±17
4	▼ 2	Realtime TTS-2闭源Inworld	1,245	±18
5	▼ 1	Simba 3.2闭源SpeechifyAI	1,237	±14
6	新	Gemini 3.8 Flash-Lite TTS闭源Google	1,235	±16
7	▼ 2	Luna TTS闭源VUI Labs	1,230	±14
8	新	Realtime TTS-2 Flash闭源Inworld	1,210	±15
9	▼ 2	Breeze TTS 2开源BreezeBlue	1,204	±16
10	▼ 1	Gemini 3.1 Flash TTS闭源Google	1,199	±12
11	▼ 1	StepAudio 2.5 TTS (Aug 2026)闭源StepFun	1,199	±17
12	▼ 4	v3 Conversational闭源ElevenLabs	1,196	±15
13	▼ 2	Sonic 3.5闭源Cartesia	1,185	±12
14	▼ 2	Lightning V3.1 Pro (Jul 2026)闭源Smallest.ai	1,175	±14
15	▼ 2	Soniox TTS Real-Time v2闭源Soniox	1,173	±12
16	▼ 1	Speech 2.8 HD闭源MiniMax	1,168	±11
17	▼ 3	Eleven v3闭源ElevenLabs	1,167	±11
18	▼ 2	Falcon 2闭源Murf AI	1,157	±15
19	▼ 2	Speech 2.8 Turbo闭源MiniMax	1,149	±11
20	▼ 2	Gradium TTS (Aug 2026)闭源Gradium	1,148	±14
21	▼ 1	Fish Audio S2.1 Pro闭源Fish Audio	1,139	±13
22	▼ 1	Step TTS 2 (Mar 2026)闭源StepFun	1,131	±13
23	▲ 1	Azure HD 2.5闭源Microsoft	1,128	±12
24	▲ 3	SpaceXAI TTS闭源SpaceXAI	1,128	±15
25	▼ 6	Async Flash v1.5闭源Async	1,128	±13
26	▼ 4	Speech 2.6 HD闭源MiniMax	1,127	±12
27	▼ 2	Fish Audio S2 Pro开源Fish Audio	1,120	±13
28	—	Simba 3.0闭源SpeechifyAI	1,118	±12
29	▼ 3	Speech 2.6 Turbo闭源MiniMax	1,114	±11
30	▼ 1	Async Pro v1.0闭源Async	1,113	±13
31	▼ 1	Speech-02-HD闭源MiniMax	1,111	±12
32	▼ 1	TTS-1 HD闭源OpenAI	1,099	±12
33	▲ 2	Chatterbox HD闭源Resemble AI	1,096	±13
34	▲ 4	Speech 2.6闭源Hithink	1,096	±13
35	▼ 2	Turbo v2.5闭源ElevenLabs	1,095	±11
36	▼ 4	Step Audio EditX (Mar 2026)开源StepFun	1,094	±13
37	▼ 3	Multilingual v2闭源ElevenLabs	1,092	±10
38	▼ 2	TTS-1闭源OpenAI	1,086	±11
39	—	Speech-02-Turbo闭源MiniMax	1,085	±12
40	▼ 3	Gemini 2.5 Flash Lite TTS闭源Google	1,081	±12
41	—	OpenAudio S1闭源Fish Audio	1,077	±11
42	▲ 1	Voxtral TTS开源Mistral	1,076	±13
43	▼ 3	Studio闭源Google	1,075	±12
44	▼ 2	Flash v2.5闭源ElevenLabs	1,074	±11
45	▼ 1	Journey闭源Google	1,073	±13
46	▲ 1	GPT-Realtime-2闭源OpenAI	1,072	±13
47	▼ 2	Maya 2 Flash闭源Maya Research	1,066	±14
48	▲ 1	Magpie-Multilingual 357M (Feb 2026)开源NVIDIA	1,063	±13
49	▲ 3	Kokoro 82M v1.0开源Kokoro	1,061	±11
50	▼ 2	T2A-01-HD闭源MiniMax	1,061	±11
51	▼ 1	Polly Generative闭源Amazon	1,061	±12
52	▼ 6	Sonic 3闭源Cartesia	1,056	±12
53	▲ 2	Coda闭源Rime	1,054	±13
54	▲ 2	Chirp 3: HD闭源Google	1,052	±12
55	▲ 3	Gemini 2.5 Flash TTS (Dec 2025)闭源Google	1,051	±12
56	▲ 4	Async Flash v1.0闭源Async	1,049	±12
57	—	Octave 2闭源Hume AI	1,049	±12
58	▼ 5	MiMo-V2.5-TTS闭源Xiaomi	1,049	±13
59	▼ 6	Maya 2 Global闭源Maya Research	1,047	±14
60	▼ 1	OpenAudio S1 Mini开源Fish Audio	1,041	±20
61	▲ 1	Maya1开源Maya Research	1,041	±12
62	▲ 4	Bland Speech v3闭源Bland AI	1,033	±13
63	—	Higgs Audio V3 TTS开源Boson AI	1,033	±13
64	▼ 3	Polly Long-Form闭源Amazon	1,032	±13
65	▲ 5	Azure Neural闭源Microsoft	1,031	±19
66	▼ 1	MAI-Voice-1闭源Microsoft	1,030	±13
67	▲ 2	Octave TTS闭源Hume AI	1,027	±12
68	▼ 1	T2A-01-Turbo闭源MiniMax	1,027	±11
69	▼ 1	Gemini 2.5 Pro (Dec 2025)闭源Google	1,027	±13
70	▲ 2	Chatterbox开源Resemble AI	1,021	±11
71	—	MiMo-V2-TTS闭源Xiaomi	1,021	±14
72	▲ 1	Lightning v3.1闭源Smallest.ai	1,020	±13
73	▲ 1	Raon SpeechLM闭源Krafton	1,002	±14
74	▲ 1	Arcana v3闭源Rime	1,000	±13
75	▲ 1	Zonos-v0.1开源Zyphra	1,000	±0
76	▲ 1	Murf Speech Gen 2闭源Murf AI	983	±12
77	▲ 2	VibeVoice 1.5B开源Microsoft	951	±14
78	▲ 2	OpenVoice v2开源OpenVoice	946	±13
79	▲ 3	Qwen3 TTS Flash闭源Alibaba	941	±14
80	▲ 3	Neuphonic TTS闭源Neuphonic	936	±14
81	▲ 3	Kugel 3闭源KugelAudio	931	±13
82	▲ 3	Qwen3 TTS闭源Alibaba	926	±13
83	▲ 4	WaveNet闭源Google	915	±12
84	▲ 2	XTTS v2开源Coqui	914	±15
85	▲ 3	Mist V2闭源Rime	906	±15
86	▲ 4	Neural2闭源Google	894	±12
87	▲ 2	StyleTTS 2开源StyleTTS	891	±15
88	▲ 3	Polly Neural闭源Amazon	890	±14
89	▲ 3	Standard闭源Google	881	±12
90	▲ 3	Noiz TTS闭源Noiz	881	±15
91	▲ 3	MetaVoice v1开源MetaVoice	845	±18
92	▲ 3	Polly Standard闭源Amazon	828	±16

其余 42 个模型

智力进步

每家厂商当时最强的模型在 Artificial Analysis 智力指数上的分数，按模型发布的月份画。

涨得最多 · Anthropic +57.2目前最高 · Anthropic 64.6最高与最低相差 18.1 分时间范围 2025年11月 – 2026年11月

64.6Anthropic60.0Google56.9OpenAI50.1Meta46.5Kimi