Обновлено: 31.05.2026 06:00 MSK

Независимый трекер открытых языковых моделей

Отслеживаем релизы открытых LLM и оцениваем их на русскоязычных бенчмарках. Объективно, прозрачно, с открытым кодом.

201 моделей отслеживается
3 источников данных
136 open-source моделей

Рейтинг моделей

Данные: Artificial Analysis (Intelligence Index) · TIGER-Lab MMLU-Pro · Arena AI

# Модель Intelligence ↓ Coding Speed t/s TTFT $/1M out License
1 Claude Opus 4.8 MAX Proprietary REASONING
Anthropic 		61.4 56.7 58 18с $25.0 Proprietary
2 GPT-5.5 MAX Proprietary REASONING
OpenAI 		60.2 59.1 56 79с $30.0 Proprietary
3 GPT-5.5 HIGH Proprietary REASONING
OpenAI 		58.9 58.5 51 19с $30.0 Proprietary
4 Claude Opus 4.7 MAX Proprietary REASONING
Anthropic 		57.3 52.5 43 11с $25.0 Proprietary
5 Gemini 3.1 Pro Preview Proprietary REASONING
Google 		57.2 55.5 123 20с $12.0 Proprietary
6 GPT-5.5 MED Proprietary REASONING
OpenAI 		56.7 56.2 48 8.0с $30.0 Proprietary
7 Qwen3.7 Max Proprietary REASONING
Alibaba 		56.6 50.1 189 2.6с $7.5 Proprietary
8 Gemini 3.5 Flash HIGH Proprietary REASONING
Google 		55.3 45.0 183 18с $9.0 Proprietary
9 Gemini 3.5 Flash MED Proprietary REASONING
Google 		54.8 43.9 174 13с $9.0 Proprietary
10 Kimi K2.6 OSS REASONING
Kimi 🤗 		53.9 47.1 44 2.3с $4.0 Modified MIT
11 MiMo-V2.5-Pro OSS REASONING
Xiaomi 🤗 		53.8 45.5 49 3.5с $0.87 Mit
12 GPT-5.3 Codex MAX Proprietary REASONING
OpenAI 		53.6 53.1 76 100с $14.0 Proprietary
13 Grok 4.3 HIGH Proprietary REASONING
xAI 		53.2 41.0 141 11с $2.5 Proprietary
14 Muse Spark Proprietary REASONING
Meta 		52.2 47.5 — — — Proprietary
15 Claude Opus 4.7 HIGH Proprietary
Anthropic 		51.8 53.1 42 1.2с $25.0 Proprietary
16 Claude Sonnet 4.6 MAX Proprietary REASONING
Anthropic 		51.7 50.9 47 105с $15.0 Proprietary
17 DeepSeek V4 Pro MAX OSS REASONING
DeepSeek 🤗 		51.5 47.5 48 1.9с $0.87 MIT
18 GLM-5.1 OSS REASONING
Z AI 🤗 		51.4 43.4 62 1.6с $4.4 Mit
19 GPT-5.5 LOW Proprietary REASONING
OpenAI 		50.8 52.1 51 1.7с $30.0 Proprietary
20 Qwen3.6 Plus Proprietary REASONING
Alibaba 		50.0 42.9 53 2.9с $3.0 Proprietary
21 DeepSeek V4 Pro HIGH OSS REASONING
DeepSeek 🤗 		49.8 43.2 46 1.8с $0.87 MIT
22 MiniMax-M2.7 OSS REASONING
MiniMax 🤗 		49.6 41.9 90 2.6с $1.2 NON-COMMERCIAL LICENSE
23 MiMo-V2.5 OSS REASONING
Xiaomi 🤗 		49.0 42.1 90 3.0с $0.28 Mit
24 GPT-5.4 mini MAX Proprietary REASONING
OpenAI 		48.9 51.5 161 7.0с $4.5 Proprietary
25 Grok 4.3 MED Proprietary REASONING
xAI 		48.8 35.1 142 7.5с $2.5 Proprietary
26 GLM-5-Turbo Proprietary REASONING
Z AI 		46.8 36.8 — — — Proprietary
27 DeepSeek V4 Flash MAX OSS REASONING
DeepSeek 🤗 		46.5 38.7 105 1.2с $0.28 MIT
28 DeepSeek V4 Flash HIGH OSS REASONING
DeepSeek 🤗 		46.0 39.8 — — $0.28 MIT
29 Qwen3.6 27B OSS REASONING
Alibaba 🤗 		45.8 36.5 57 3.8с $3.6 Apache 2.0
30 Qwen3.5 397B A17B OSS REASONING
Alibaba 🤗 		45.0 41.3 53 2.6с $3.6 Apache 2.0
31 MiMo-V2-Omni-0327 Proprietary REASONING
Xiaomi 		44.9 36.9 92 3.8с $2.0 Proprietary
32 Claude Sonnet 4.6 HIGH Proprietary
Anthropic 		44.4 46.4 42 1.2с $15.0 Proprietary
33 GPT-5.4 nano MAX Proprietary REASONING
OpenAI 		44.0 43.9 154 3.8с $1.2 Proprietary
34 Grok 4.3 LOW Proprietary REASONING
xAI 		43.9 31.6 109 4.3с $2.5 Proprietary
35 GLM-5.1 OSS
Z AI 🤗 		43.8 35.8 49 1.8с $4.4 Mit
36 Qwen3.6 35B A3B OSS REASONING
Alibaba 🤗 		43.5 35.1 174 2.4с $1.5 Apache 2.0
37 MiMo-V2-Omni Proprietary REASONING
Xiaomi 		43.4 35.5 90 3.8с free Proprietary
38 Gemini 3.5 Flash LOW Proprietary
Google 		43.3 47.1 169 0.9с $9.0 Proprietary
39 Kimi K2.6 OSS
Kimi 🤗 		42.9 38.4 42 2.5с $4.0 Modified MIT
40 GLM 5V Turbo Proprietary REASONING
Z AI 		42.9 36.2 — — — Proprietary
41 Claude Sonnet 4.6 LOW Proprietary
Anthropic 		42.6 43.0 42 1.3с $15.0 Proprietary
42 Hy3-preview OSS REASONING
Tencent 🤗 		41.9 36.5 99 3.9с $0.43 TENCENT HY COMMUNITY LICENSE AGREEMENT
43 GPT-5.5 Instant MAY 2026 Proprietary REASONING
OpenAI 		41.8 45.1 — — $30.0 Proprietary
44 Qwen3.5 122B A10B OSS REASONING
Alibaba 🤗 		41.6 34.7 139 2.5с $3.2 Apache 2.0
45 MiMo-V2-Flash FEB 2026 OSS REASONING
Xiaomi 🤗 		41.5 33.5 131 2.1с $0.30 MIT
46 GPT-5.5 Proprietary
OpenAI 		40.9 48.6 50 1.0с $30.0 Proprietary
47 Qwen3.5 397B A17B OSS
Alibaba 🤗 		40.1 37.4 53 2.6с $3.6 Apache 2.0
48 DeepSeek V4 Pro OSS
DeepSeek 🤗 		39.3 38.4 52 2.0с $0.87 Mit
49 Gemma 4 31B OSS REASONING
Google 🤗 		39.2 38.7 35 1.1с free Apache 2.0
50 Mistral Medium 3.5 OSS REASONING
Mistral 🤗 		39.2 35.4 148 1.8с $7.5 Other
51 Qwen3.5 Omni Plus Proprietary
Alibaba 		38.6 27.6 55 2.5с $4.8 Proprietary
52 Step 3.5 Flash 2603 Proprietary REASONING
StepFun 		38.5 34.6 159 1.2с free Proprietary
53 Ring-2.6-1T OSS REASONING
InclusionAI 🤗 		38.5 33.3 122 3.2с $2.5 MIT
54 o3 Proprietary REASONING
OpenAI 		38.4 38.4 126 5.5с $8.0 Proprietary
55 GPT-5.4 nano MED Proprietary REASONING
OpenAI 		38.1 35.0 146 2.9с $1.2 Proprietary
56 GPT-5.4 mini MED Proprietary REASONING
OpenAI 		37.7 37.5 157 4.1с $4.5 Proprietary
57 Command A+ OSS REASONING
Cohere 🤗 		37.2 29.3 223 0.3с free Apache 2.0
58 Qwen3.6 27B OSS
Alibaba 🤗 		37.1 26.6 56 3.9с $3.6 Apache 2.0
59 Claude 4.5 Haiku Proprietary REASONING
Anthropic 		37.1 32.6 94 24с $5.0 Proprietary
60 DeepSeek V4 Flash OSS
DeepSeek 🤗 		36.5 35.1 106 1.4с $0.28 Mit
61 JT-35B-Flash Proprietary
China Mobile 		36.1 28.9 — — — Proprietary
62 NVIDIA Nemotron 3 Super 120B A12B OSS REASONING
NVIDIA 🤗 		36.0 31.2 181 1.8с $0.75 Nvidia Nemotron Open Model License
63 Qwen3.5 122B A10B OSS
Alibaba 🤗 		35.9 31.6 162 2.5с $3.2 Apache 2.0
64 Nova 2.0 Pro Preview MED Proprietary REASONING
Amazon 		35.7 30.4 114 13с $10.0 Proprietary
65 MiMo-V2.5-Pro OSS
Xiaomi 🤗 		35.6 36.8 48 3.4с $2.7 Mit
66 Gemini 2.5 Pro Proprietary REASONING
Google 		34.6 31.9 123 23с $10.0 Proprietary
67 Nova 2.0 Lite HIGH Proprietary REASONING
Amazon 		34.5 23.4 148 14с $2.5 Proprietary
68 Hy3-preview OSS
Tencent 🤗 		33.7 34.3 89 4.0с $0.43 TENCENT HY COMMUNITY LICENSE AGREEMENT
69 Ling-2.6-1T OSS
InclusionAI 🤗 		33.6 33.0 — — $2.5 Mit
70 Doubao Seed Code Proprietary REASONING
ByteDance Seed 		33.5 31.3 — — — Proprietary
71 Gemini 3.1 Flash-Lite Proprietary REASONING
Google 		33.5 30.1 256 5.6с $1.5 Proprietary
72 gpt-oss-120b HIGH OSS REASONING
OpenAI 🤗 		33.3 28.6 322 0.9с $0.60 Apache 2.0
73 Mercury 2 Proprietary REASONING
Inception 		32.8 30.6 744 2.5с $0.75 Proprietary
74 Qwen3.5 9B OSS REASONING
Alibaba 🤗 		32.4 25.3 68 2.4с $0.15 Apache 2.0
75 Gemma 4 31B OSS
Google 🤗 		32.3 33.9 17 1.4с $0.40 Apache 2.0
76 K-EXAONE OSS REASONING
LG AI Research 🤗 		32.1 27.0 — — — K Exaone
77 Trinity Large Thinking OSS REASONING
Arcee AI 🤗 		31.9 27.2 157 1.2с $0.88 Apache 2.0
78 Nova 2.0 Pro Preview LOW Proprietary REASONING
Amazon 		31.9 24.5 118 10с $10.0 Proprietary
79 Qwen3.6 35B A3B OSS
Alibaba 🤗 		31.5 17.6 179 2.6с $2.2 Apache 2.0
80 Gemma 4 26B A4B OSS REASONING
Google 🤗 		31.2 22.4 — — $0.40 Apache 2.0
81 Grok 4.3 Proprietary
xAI 		31.0 25.1 110 0.6с $2.5 Proprietary
82 Claude 4.5 Haiku Proprietary
Anthropic 		31.0 29.6 90 0.8с $5.0 Proprietary
83 Qwen3.5 35B A3B OSS
Alibaba 🤗 		30.7 16.8 155 2.1с $2.0 Apache 2.0
84 MiMo-V2-Flash OSS
Xiaomi 🤗 		30.3 25.8 130 2.1с $0.30 MIT
85 EXAONE 4.5 33B OSS REASONING
LG AI Research 🤗 		30.2 23.0 — — — EXAONE AI Model License Agreement 1.2 - NC
86 Nova 2.0 Lite MED Proprietary REASONING
Amazon 		29.7 23.9 147 21с $2.5 Proprietary
87 ERNIE 5.0 Thinking Preview Proprietary REASONING
Baidu 		29.1 29.2 — — — Proprietary
88 Nemotron Cascade 2 30B A3B OSS REASONING
NVIDIA 🤗 		28.4 25.8 — — — Nvidia Open Model License
89 Qwen3 Coder Next OSS
Alibaba 🤗 		28.3 22.9 106 1.6с $1.2 Apache 2.0
90 Nova 2.0 Omni MED Proprietary REASONING
Amazon 		28.0 15.1 — — $2.5 Proprietary
91 Mistral Small 4 OSS REASONING
Mistral 🤗 		27.8 24.3 181 0.7с $0.60 Apache 2.0
92 Qwen3.5 9B OSS
Alibaba 🤗 		27.3 21.4 — — — Apache 2.0
93 Qwen3.5 4B OSS REASONING
Alibaba 🤗 		27.1 17.5 192 0.4с $0.15 Apache 2.0
94 Gemma 4 26B A4B OSS
Google 🤗 		27.1 29.1 78 1.6с $0.40 Apache 2.0
95 Magistral Medium 1.2 Proprietary REASONING
Mistral 		27.1 21.7 38 1.9с $5.0 Proprietary
96 Qwen3 Next 80B A3B OSS REASONING
Alibaba 🤗 		26.7 19.5 136 2.3с $6.0 Apache 2.0
97 Ling 2.6 Flash OSS
InclusionAI 🤗 		26.2 23.2 — — $0.30 Mit
98 Qwen3.5 Omni Flash Proprietary
Alibaba 		25.9 14.0 241 1.9с $0.80 Proprietary
99 Solar Pro 3 Proprietary REASONING
Upstage 		25.9 13.3 — — — Proprietary
100 JT-MINI Proprietary
China Mobile 		25.4 21.2 — — — Proprietary
101 Nova 2.0 Lite LOW Proprietary REASONING
Amazon 		24.6 13.6 152 8.5с $2.5 Proprietary
102 gpt-oss-20B HIGH OSS REASONING
OpenAI 🤗 		24.5 18.5 239 0.7с $0.20 Apache 2.0
103 gpt-oss-120b LOW OSS REASONING
OpenAI 🤗 		24.5 15.5 345 0.9с $0.60 Apache 2.0
104 GPT-5.4 nano Proprietary
OpenAI 		24.4 27.9 148 0.6с $1.2 Proprietary
105 NVIDIA Nemotron 3 Nano 30B A3B OSS REASONING
NVIDIA 🤗 		24.3 19.0 132 2.0с $0.22 Nvidia Nemotron Open Model License
106 LongCat Flash Lite OSS
LongCat 🤗 		23.9 16.5 91 6.3с free MIT
107 K-EXAONE OSS
LG AI Research 🤗 		23.4 13.5 — — — K Exaone
108 GPT-5.4 mini Proprietary
OpenAI 		23.3 25.3 143 0.7с $4.5 Proprietary
109 Nova 2.0 Omni LOW Proprietary REASONING
Amazon 		23.2 13.9 — — $2.5 Proprietary
110 Nova 2.0 Pro Preview Proprietary
Amazon 		23.1 20.5 119 1.1с $10.0 Proprietary
111 Mi:dm K 2.5 Pro Proprietary REASONING
Korea Telecom 		23.1 12.6 — — — Proprietary
112 Mistral Large 3 OSS
Mistral 🤗 		22.8 22.7 52 1.1с $1.5 Apache 2.0
113 Qwen3.5 4B OSS
Alibaba 🤗 		22.6 13.7 198 0.4с $0.15 Apache 2.0
114 INTELLECT-3 OSS REASONING
Prime Intellect 🤗 		22.2 19.1 — — — MIT
115 Devstral 2 OSS
Mistral 🤗 		22.0 23.7 64 1.2с free Modified MIT License
116 Solar Open 100B OSS REASONING
Upstage 🤗 		21.7 10.5 — — — Upstage Solar License
117 Nemotron 3 Nano Omni 30B A3B Reasoning OSS REASONING
NVIDIA 🤗 		21.4 14.8 299 1.0с $0.30 NVIDIA Open Model License Agreement
118 gpt-oss-20B LOW OSS REASONING
OpenAI 🤗 		20.8 14.4 242 0.8с $0.20 Apache 2.0
119 Qwen3 Next 80B A3B Instruct OSS
Alibaba 🤗 		20.1 15.3 149 2.3с $2.0 Apache 2.0
120 Devstral Small 2 OSS
Mistral 🤗 		19.5 20.7 68 1.1с free Apache 2.0
121 Motif-2-12.7B-Reasoning Proprietary REASONING
Motif Technologies 		19.1 11.9 — — — Proprietary
122 Nova Premier Proprietary
Amazon 		19.0 13.8 35 2.9с $12.5 Proprietary
123 Gemma 4 E4B OSS REASONING
Google 🤗 		18.8 13.7 — — — Apache 2.0
124 Llama Nemotron Super 49B v1.5 OSS REASONING
NVIDIA 🤗 		18.7 15.2 47 1.3с $0.40 NVIDIA Open Model License Agreement
125 Mistral Small 4 OSS
Mistral 🤗 		18.6 16.4 159 0.7с $0.60 Apache 2.0
126 Llama 4 Maverick OSS
Meta 🤗 		18.4 15.6 111 1.0с $0.85 LLAMA 4 COMMUNITY LICENSE AGREEMENT
127 Sarvam 105B HIGH OSS REASONING
Sarvam 🤗 		18.2 9.8 97 2.1с $0.17 Apache 2.0
128 Magistral Small 1.2 OSS REASONING
Mistral 🤗 		18.2 14.8 108 0.8с $1.5 Apache 2.0
129 Nova 2.0 Lite Proprietary
Amazon 		18.0 12.5 142 1.3с $2.5 Proprietary
130 MiniCPM5-1B OSS
OpenBMB 🤗 		17.9 0.5 — — — apache-2.0
131 Llama 3.1 Instruct 405B OSS
Meta 🤗 		17.4 14.5 40 2.3с $6.5 LLAMA 3.1 COMMUNITY LICENSE AGREEMENT
132 EXAONE 4.0 32B OSS REASONING
LG AI Research 🤗 		16.7 14.0 — — — EXAONE AI Model License Agreement 1.2 - NC
133 Nova 2.0 Omni Proprietary
Amazon 		16.6 13.8 — — $2.5 Proprietary
134 Qwen3.5 2B OSS REASONING
Alibaba 🤗 		16.3 3.5 — — $0.10 Apache 2.0
135 Nanbeige4.1-3B OSS REASONING
Nanbeige 🤗 		16.1 8.9 — — — Apache 2.0
136 Ministral 3 14B OSS
Mistral 🤗 		16.0 10.9 77 0.8с $0.20 Apache 2.0
137 Falcon-H1R-7B OSS REASONING
TII UAE 🤗 		15.8 9.8 — — — Falcon Mamba 7B TII License Version 1.0
138 Qwen3 Omni 30B A3B OSS REASONING
Alibaba 🤗 		15.6 12.7 89 2.0с $0.97 Apache 2.0
139 Step3 VL 10B OSS REASONING
StepFun 🤗 		15.5 13.9 — — — Apache 2.0
140 Gemma 4 E2B OSS REASONING
Google 🤗 		15.2 9.0 — — — Apache 2.0
141 ERNIE 4.5 300B A47B OSS
Baidu 🤗 		15.0 14.5 25 3.5с $1.1 Apache 2.0
142 Llama 3.1 Nemotron Ultra 253B v1 OSS REASONING
NVIDIA 🤗 		15.0 13.1 52 2.4с $1.8 NVIDIA Open Model License Agreement
143 NVIDIA Nemotron Nano 12B v2 VL OSS REASONING
NVIDIA 🤗 		14.9 11.8 — — $0.60 Nvidia Open Model License
144 Solar Pro 2 Proprietary REASONING
Upstage 		14.9 12.1 — — — Proprietary
145 Gemma 4 E4B OSS
Google 🤗 		14.8 6.4 — — — Apache 2.0
146 Ministral 3 8B OSS
Mistral 🤗 		14.8 10.0 98 0.7с $0.15 Apache 2.0
147 NVIDIA Nemotron Nano 9B V2 OSS REASONING
NVIDIA 🤗 		14.8 8.3 123 0.7с $0.16 NVIDIA Open Model License Agreement
148 Qwen3.5 2B OSS
Alibaba 🤗 		14.7 4.9 247 0.4с $0.10 Apache 2.0
149 NVIDIA Nemotron 3 Nano 4B OSS REASONING
NVIDIA 🤗 		14.7 10.0 — — — Nvidia Nemotron Open Model License
150 Granite 4.1 30B OSS
IBM 🤗 		14.7 10.1 — — — Apache 2.0
151 Llama Nemotron Super 49B v1.5 OSS
NVIDIA 🤗 		14.6 10.5 48 1.3с $0.40 NVIDIA Open Model License Agreement
152 Llama 3.3 Instruct 70B OSS
Meta 🤗 		14.5 10.7 80 1.6с $0.71 LLAMA 3.3 COMMUNITY LICENSE AGREEMENT
153 Kimi Linear 48B A3B Instruct OSS
Kimi 🤗 		14.4 14.2 — — — MIT
154 Ring-flash-2.0 OSS REASONING
InclusionAI 🤗 		14.0 10.6 — — $0.57 MIT
155 Solar Pro 2 Proprietary
Upstage 		13.6 11.3 — — — Proprietary
156 Command A OSS
Cohere 🤗 		13.5 9.9 55 1.8с $10.0 CC-BY-NC 4.0 License with Acceptable Use Addendum
157 Llama 4 Scout OSS
Meta 🤗 		13.5 6.7 105 0.9с $0.66 LLAMA 4 COMMUNITY LICENSE AGREEMENT
158 Llama 3.1 Nemotron Instruct 70B OSS
NVIDIA 🤗 		13.4 10.8 292 0.5с $1.2 LLAMA 3.1 COMMUNITY LICENSE AGREEMENT
159 NVIDIA Nemotron 3 Nano 30B A3B OSS
NVIDIA 🤗 		13.2 15.8 83 0.4с $0.20 Nvidia Nemotron Open Model License
160 NVIDIA Nemotron Nano 9B V2 OSS
NVIDIA 🤗 		13.2 7.5 142 1.1с $0.20 NVIDIA Open Model License Agreement
161 MiniCPM-V 4.6 1.3B OSS
OpenBMB 🤗 		12.7 0.7 — — — Apache 2.0
162 Granite 4.1 8B OSS
IBM 🤗 		12.4 7.2 108 0.8с $0.10 Apache 2.0
163 Sarvam 30B HIGH OSS REASONING
Sarvam 🤗 		12.3 7.9 163 1.9с $0.11 Apache 2.0
164 Gemma 4 E2B OSS
Google 🤗 		12.1 8.3 — — — Apache 2.0
165 R1 1776 OSS REASONING
Perplexity 🤗 		12.0 — — — — MIT
166 Llama 3.2 Instruct 90B VISION OSS
Meta 🤗 		11.9 — 58 1.2с $1.4 LLAMA 3.2 COMMUNITY LICENSE AGREEMENT
167 EXAONE 4.0 32B OSS
LG AI Research 🤗 		11.7 9.4 — — — EXAONE AI Model License Agreement 1.2 - NC
168 Ministral 3 3B OSS
Mistral 🤗 		11.2 4.8 199 0.5с $0.10 Apache 2.0
169 Jamba 1.7 Large OSS
AI21 Labs 🤗 		10.9 7.8 62 1.7с $8.0 Jamba Open Model License Agreement
170 Granite 4.0 H Small OSS
IBM 🤗 		10.8 8.5 357 10с $0.25 Apache 2.0
171 Qwen3 Omni 30B A3B Instruct OSS
Alibaba 🤗 		10.7 7.2 96 2.1с $0.97 Apache 2.0
172 Qwen3.5 0.8B OSS REASONING
Alibaba 🤗 		10.5 0.0 — — $0.050 Apache 2.0
173 LFM2 24B A2B OSS
Liquid AI 🤗 		10.5 3.6 120 0.6с $0.12 lfm 1.0
174 Phi-4 OSS
Microsoft 🤗 		10.4 11.2 38 2.0с $0.50 MIT
175 Nova Micro Proprietary
Amazon 		10.3 4.1 290 0.9с $0.14 Proprietary
176 NVIDIA Nemotron Nano 12B v2 VL OSS
NVIDIA 🤗 		10.1 5.9 227 1.1с $0.60 Nvidia Open Model License
177 Phi-4 Multimodal Instruct OSS
Microsoft 🤗 		10.0 — 16 1.1с free MIT
178 Qwen3.5 0.8B OSS
Alibaba 🤗 		9.9 1.0 74 0.4с $0.050 Apache 2.0
179 Jamba Reasoning 3B OSS REASONING
AI21 Labs 🤗 		9.6 2.5 — — — Apache 2.0
180 Reka Flash 3 OSS REASONING
Reka AI 🤗 		9.5 8.9 — — $0.80 Apache 2.0
181 Ling-mini-2.0 OSS
InclusionAI 🤗 		9.2 5.0 — — — MIT
182 Llama 3.2 Instruct 11B VISION OSS
Meta 🤗 		8.7 4.2 53 0.7с $0.24 LLAMA 3.2 COMMUNITY LICENSE AGREEMENT
183 Granite 4.1 3B OSS
IBM 🤗 		8.5 5.5 — — — Apache 2.0
184 Phi-4 Mini Instruct OSS
Microsoft 🤗 		8.4 3.6 — — free MIT
185 Exaone 4.0 1.2B OSS REASONING
LG AI Research 🤗 		8.3 3.1 — — — EXAONE AI Model License Agreement 1.2 - NC
186 LFM2.5-1.2B-Thinking OSS REASONING
Liquid AI 🤗 		8.1 1.4 — — — lfm 1.0
187 Exaone 4.0 1.2B OSS
LG AI Research 🤗 		8.1 2.5 — — — EXAONE AI Model License Agreement 1.2 - NC
188 Jamba 1.7 Mini OSS
AI21 Labs 🤗 		8.1 3.1 — — — Jamba Open Model License Agreement
189 LFM2.5-1.2B-Instruct OSS
Liquid AI 🤗 		8.0 0.8 — — free lfm 1.0
190 LFM2 2.6B OSS
Liquid AI 🤗 		8.0 1.4 — — free lfm 1.0
191 Granite 4.0 H 1B OSS
IBM 🤗 		8.0 2.7 — — — Apache 2.0
192 Gemma 3 270M OSS
Google 🤗 		7.7 0.0 — — — Gemma license
193 Granite 4.0 Micro OSS
IBM 🤗 		7.7 5.0 — — — Apache 2.0
194 Apertus 70B Instruct OSS
Swiss AI Initiative 🤗 		7.7 1.9 — — $2.9 Apache 2.0
195 Granite 4.0 1B OSS
IBM 🤗 		7.3 2.9 — — — Apache 2.0
196 LFM2 8B A1B OSS
Liquid AI 🤗 		7.0 2.3 — — free lfm 1.0
197 LFM2.5-VL-1.6B OSS
Liquid AI 🤗 		6.2 1.0 — — free lfm 1.0
198 Granite 4.0 350M OSS
IBM 🤗 		6.1 0.3 — — — Apache 2.0
199 Apertus 8B Instruct OSS
Swiss AI Initiative 🤗 		5.9 1.4 — — $0.20 Apache 2.0
200 Granite 4.0 H 350M OSS
IBM 🤗 		5.4 0.6 — — — Apache 2.0
201 Tiny Aya Global OSS
Cohere 🤗 		4.7 1.2 — — free Cc By Nc 4.0
Intelligence — Artificial Analysis Blended (MMLU-Pro, GPQA, HLE, IFBench, etc.) Coding — AA Coding Index (LiveCodeBench, SciCode, Aider) Speed — медианная скорость генерации, токенов/сек TTFT — Time to First Token, секунды (медиана) $/1M out — цена за 1M output-токенов (USD)

Рейтинг Arena AI

Живой рейтинг на основе слепых попарных сравнений пользователями · 250 open-source моделей · Данные: arena.ai

# Модель Score CI Голоса
1 glm-5.1
Z.ai		 1474 ±6 13K
2 mimo-v2.5-pro
Xiaomi		 1465 ±6 15K
3 kimi-k2.6
Moonshot		 1462 ±6 15K
4 deepseek-v4-pro-thinking
DeepSeek		 1458 ±6 15K
5 glm-5
Z.ai		 1457 ±5 21K
6 deepseek-v4-pro
DeepSeek		 1454 ±6 16K
7 gemma-4-31b
Google		 1452 ±8 5K
8 kimi-k2.5-thinking
Moonshot		 1449 ±4 36K
9 qwen3.5-397b-a17b
Alibaba		 1445 ±4 31K
10 glm-4.7
Z.ai		 1443 ±6 12K
11 gemma-4-26b-a4b
Google		 1439 ±8 5K
12 deepseek-v4-flash-thinking
DeepSeek		 1437 ±6 16K
13 mimo-v2.5
Xiaomi		 1434 ±6 15K
14 deepseek-v4-flash
DeepSeek		 1433 ±6 16K
15 kimi-k2.5-instant
Moonshot		 1432 ±7 8K
16 kimi-k2-thinking-turbo
Moonshot		 1430 ±3 60K
17 glm-4.6
Z.ai		 1426 ±4 35K
18 deepseek-v3.2-exp-thinking
DeepSeek		 1425 ±7 9K
19 deepseek-v3.2
DeepSeek		 1424 ±4 46K
20 deepseek-v3.2-exp
DeepSeek		 1423 ±6 11K
21 qwen3-235b-a22b-instruct-2507
Alibaba		 1423 ±3 95K
22 deepseek-r1-0528
DeepSeek		 1422 ±6 18K
23 deepseek-v3.2-thinking
DeepSeek		 1422 ±4 40K
24 kimi-k2-0905-preview
Moonshot		 1418 ±6 11K
25 deepseek-v3.1
DeepSeek		 1418 ±6 14K
26 deepseek-v3.1-terminus-thinking
DeepSeek		 1418 ±10 3K
27 kimi-k2-0711-preview
Moonshot		 1417 ±5 27K
28 qwen3.5-122b-a10b
Alibaba		 1417 ±4 26K
29 deepseek-v3.1-thinking
DeepSeek		 1417 ±7 11K
30 deepseek-v3.1-terminus
DeepSeek		 1416 ±10 3K
205 open-source моделей

Свежие релизы

по дате выхода
Claude Opus 4.8 MAX
Anthropic
Proprietary
Intel 61.4 28 май 2026 · 58 t/s · $25.0/1M
MiniCPM5-1B
OpenBMB
OSS
Intel 17.9 25 май 2026
Подробнее →
Command A+
Cohere
OSS
Intel 37.2 20 май 2026 · 223 t/s · free/1M
Подробнее →
Qwen3.7 Max
Alibaba
Proprietary
Intel 56.6 19 май 2026 · 189 t/s · $7.5/1M
Gemini 3.5 Flash HIGH
Google
Proprietary
Intel 55.3 19 май 2026 · 183 t/s · $9.0/1M
Gemini 3.5 Flash MED
Google
Proprietary
Intel 54.8 19 май 2026 · 174 t/s · $9.0/1M
Gemini 3.5 Flash LOW
Google
Proprietary
Intel 43.3 19 май 2026 · 169 t/s · $9.0/1M
JT-35B-Flash
China Mobile
Proprietary
Intel 36.1 14 май 2026

Лучшие OSS

Intelligence Index
Kimi K2.6
Kimi
OSS
Intel 53.9 20 апр 2026 · 44 t/s · $4.0/1M
Подробнее →
MiMo-V2.5-Pro
Xiaomi
OSS
Intel 53.8 22 апр 2026 · 49 t/s · $0.87/1M
Подробнее →
DeepSeek V4 Pro MAX
DeepSeek
OSS
Intel 51.5 24 апр 2026 · 48 t/s · $0.87/1M
Подробнее →
GLM-5.1
Z AI
OSS
Intel 51.4 7 апр 2026 · 62 t/s · $4.4/1M
Подробнее →
DeepSeek V4 Pro HIGH
DeepSeek
OSS
Intel 49.8 24 апр 2026 · 46 t/s · $0.87/1M
Подробнее →
MiniMax-M2.7
MiniMax
OSS
Intel 49.6 18 мар 2026 · 90 t/s · $1.2/1M
Подробнее →
MiMo-V2.5
Xiaomi
OSS
Intel 49.0 22 апр 2026 · 90 t/s · $0.28/1M
Подробнее →
DeepSeek V4 Flash MAX
DeepSeek
OSS
Intel 46.5 24 апр 2026 · 105 t/s · $0.28/1M
Подробнее →

О проекте

LLM Bench — независимый open-source проект, отслеживающий экосистему языковых моделей. Основной рейтинг строится на данных Arena AI — крупнейшей платформы слепых попарных сравнений моделей пользователями. Дополнительно подключаются академические бенчмарки (MMLU-Pro) и данные Hugging Face.

Данные агрегируются из открытых источников: Arena AI, TIGER-Lab MMLU-Pro и Hugging Face API. Обновление происходит автоматически ежедневно. Все результаты проверяемы — ссылки ведут на оригинальные источники.

Проект поддерживается сообществом и не аффилирован ни с одним разработчиком моделей. Исходный код, данные и методология полностью открыты — вклад приветствуется через GitHub.

Методология

Arena AI (основной)
Рейтинг на основе попарных слепых сравнений пользователями. 7 категорий: общий, код, зрение, генерация изображений, редактирование, видео, поиск. >300K голосов. Данные: arena.ai.
MMLU-Pro
12 032 вопроса повышенной сложности из 14 предметных областей. Расширенная версия MMLU с 10 вариантами ответов. Дополнительная метрика. Данные: TIGER-Lab.
Агрегация
Arena AI — основной источник ранжирования. MMLU-Pro подключается как дополнительный бенчмарк при наличии совпадений. Данные обновляются ежедневно. По умолчанию показываются только open-source модели.