LLM

LLM Ranking: 257 Language Models on One Scale

Which language model is the most capable overall? 257 LLMs ranked by the Epoch Capabilities Index, plus the best open-weights models, the best model of each organization and the leaders on 15 benchmarks.

By 10 min read
Which language model is the most capable overall? 257 LLMs ranked by the Epoch Capabilities Index, plus the best open-weights models, the best model of each organization and the leaders on 15 benchmarks.

Single benchmarks disagree: one model leads in mathematics, another in coding, a third in long tasks. A ranking needs one scale. We use the Epoch Capabilities Index, which combines the results of more than 50 benchmarks into one number per model — an index, not a percentage, scaled so that Claude 3.5 Sonnet stands at 130 and GPT-5 at 150.

The ranking covers the 257 models for which the index exists. We did not test any model ourselves and changed no value; our part is the ranking views below. The margins at the top are small: differences of one or two index points are within the uncertainty of the index.

The ranking

Overall: all 257 models

Rank by Epoch Capabilities Index
#ModelReleasedWeightsECI
1Claude Opus 5.5 AnthropicSep 22, 2026closed167.3
2GPT-6 Astra OpenAISep 3, 2026closed166.4
3GPT-6.1 Sol OpenAISep 29, 2026closed166.1
4Claude Sonnet 5.5 AnthropicSep 28, 2026closed165.0
5Claude Fable 5.1 AnthropicSep 1, 2026closed164.7
6Claude Opus 5 AnthropicJul 24, 2026closed162.8
7GPT-6 Sol OpenAISep 22, 2026closed162.7
8GPT-5.5 Pro OpenAIApr 23, 2026closed162.1
9Claude Fable 5 AnthropicJun 9, 2026closed162.1
10GPT-5.6 Sol OpenAIJul 9, 2026closed161.7
11GPT-5.6 Terra OpenAIJul 9, 2026closed159.6
12GPT-5.5 OpenAIApr 23, 2026closed159.1
13GPT-5.4 Pro OpenAIMar 5, 2026closed158.9
14Claude Opus 4.8 AnthropicMay 28, 2026closed158.2
15Kimi K3 MoonshotJul 16, 2026open157.4
16Gemini 3.7 Flash Google DeepMindAug 13, 2026closed157.3
17GPT-5.3 Codex OpenAIFeb 5, 2026closed156.8
18GPT-5.4 OpenAIMar 5, 2026closed156.8
19Muse Spark 1.3 Meta AISep 2, 2026closed156.8
20Gemini 3.8 Flash Google DeepMindSep 2, 2026closed156.7
21GPT-5.6 Luna OpenAIJul 9, 2026closed156.4
22Qwen 3.8 Max AlibabaAug 2, 2026closed156.4
23Grok 4.6 xAIAug 12, 2026closed156.4
24GPT-6 Luna OpenAISep 22, 2026closed156.3
25Claude Opus 4.7 AnthropicApr 16, 2026closed156.2
26Claude Sonnet 5 AnthropicJun 30, 2026closed156.2
27GLM-5.3 Z.ai (Zhipu AI)Aug 14, 2026closed155.6
28GPT-5.2 Pro OpenAIDec 11, 2025closed155.4
29DeepSeek V4 Pro 0813 DeepSeekAug 13, 2026open155.3
30Claude Opus 4.6 AnthropicFeb 5, 2026closed155.2
31Qwen3.8 Max (0902) AlibabaSep 1, 2026closed155.1
32Muse Spark 1.2 Meta AIAug 5, 2026closed154.9
33DeepSeek V4.1 Flash DeepSeekSep 9, 2026open154.9
34Gemini 3.1 Pro Google DeepMindFeb 19, 2026closed154.8
35Gemini 3.5 Flash Google DeepMindMay 19, 2026closed154.5
36DeepSeek V4 Flash 0731 DeepSeekJul 31, 2026open154.5
37Muse Spark 1.1 Meta AIJul 9, 2026closed154.2
38Gemini 3.6 Flash Google DeepMindJul 21, 2026closed154.2
39Grok 4.5 xAIJul 8, 2026closed153.9
40Qwen3.7-Max AlibabaMay 19, 2026closed153.7
41Grok 4.7 xAISep 21, 2026closed153.5
42GPT-5.2 OpenAIDec 11, 2025closed153.4
43Gemini 3 Pro Google DeepMindNov 18, 2025closed152.9
44Claude Sonnet 4.6 AnthropicFeb 17, 2026closed152.2
45Grok 4.20 xAIFeb 17, 2026closed152.0
46Muse Spark Meta AIApr 8, 2026closed152.0
47GLM-5.3-Flash Z.ai (Zhipu AI)Aug 20, 2026open151.9
48Gemini 3 Flash Google DeepMindDec 17, 2025closed151.8
49GLM-5.2 Z.ai (Zhipu AI)Jun 16, 2026open151.8
50Kimi K2.6 MoonshotApr 20, 2026open151.1
51GPT-5 Pro OpenAIOct 7, 2025closed150.3
52Inkling-Small Thinking MachinesJul 15, 2026open150.2
53Claude Opus 4.5 AnthropicNov 24, 2025closed150.1
54GPT-5 OpenAIAug 7, 2025closed150.0
55Kimi K2.7 Code MoonshotJun 12, 2026open150.0
56GLM-5.1 Z.ai (Zhipu AI)Apr 7, 2026open149.8
57GPT-5.1 OpenAINov 13, 2025closed149.6
58Qwen 3.8 27B AlibabaAug 14, 2026open149.4
59Grok 4.3 Beta xAIApr 17, 2026closed149.2
60Qwen 3.6 Max (Preview) AlibabaApr 20, 2026–149.2
61DeepSeek-V4-Pro DeepSeekApr 24, 2026open149.1
62GPT-5.4 Mini OpenAIMar 17, 2026closed148.8
63Inkling Thinking MachinesJul 15, 2026open148.5
64Kimi K2.5 MoonshotJan 27, 2026open148.0
65Qwen 3.6 Plus AlibabaMar 31, 2026closed147.7
66o3-pro OpenAIJun 10, 2025closed147.4
67Qwen3.7-Plus AlibabaJun 2, 2026closed147.4
68o3 OpenAIApr 16, 2025closed146.9
69MiniMax-M3 MiniMaxJun 1, 2026open146.9
70Claude Sonnet 4.5 AnthropicSep 29, 2025closed146.8
71Qwen 3.5 Plus (hosted 397B-A17B) AlibabaFeb 16, 2026–146.8
72MiniMax-M2.5 MiniMaxFeb 12, 2026open146.7
73Qwen3.5 397B-A17B AlibabaFeb 13, 2026open146.7
74Qwen3.6 27B AlibabaApr 22, 2026open146.5
75Grok 4 xAIJul 9, 2025closed146.4
76DeepSeek-V3.2 DeepSeekDec 1, 2025open146.3
77Nemotron 3 Ultra NvidiaJun 4, 2026open146.2
78DeepSeek-V4-Flash DeepSeekApr 24, 2026open146.1
79Kimi K2 Thinking MoonshotNov 6, 2025open146.0
80GLM-5 Z.ai (Zhipu AI)Feb 11, 2026open145.8
81GPT-5.4 Nano OpenAIMar 17, 2026closed145.8
82MiniMax-M2.7 MiniMaxMar 18, 2026open145.8
83o4-mini OpenAIApr 16, 2025closed145.6
84GPT-5 mini OpenAIAug 7, 2025closed145.5
85Gemini 2.5 Pro (Jun 2025) Google DeepMindJun 5, 2025closed145.3
86Gemini 3.5 Flash-Lite Google DeepMindJul 21, 2026closed145.1
87DeepSeek-V3.2-Exp DeepSeekSep 29, 2025open145.0
88Qwen3.7 Flash AlibabaJul 27, 2026closed144.6
89Gemini 3.1 Flash-Lite GoogleMar 3, 2026closed144.5
90Gemini 2.5 Pro (Mar 2025) Google DeepMindMar 25, 2025closed144.2
91Grok 4 Fast xAISep 19, 2025closed144.2
92Claude Opus 4.1 AnthropicAug 5, 2025closed144.1
93Qwen 3.5 Flash (hosted 35B-A3B) AlibabaFeb 25, 2026–144.0
94Qwen 3.6 35B-A3B AlibabaApr 14, 2026open143.9
95Qwen3-235B-A22B-Thinking (Jul 2025) AlibabaJul 25, 2025open143.8
96GLM-4.7 Z.ai (Zhipu AI)Dec 22, 2025open143.5
97Qwen 3.6 Flash AlibabaApr 27, 2026–143.3
98Gemini 2.5 Flash (Sep 2025) Google DeepMindSep 25, 2025closed143.0
99Claude Opus 4 AnthropicMay 22, 2025closed142.7
100Gemma 4 31B IT Google DeepMindApr 2, 2026open142.7
101Gemini 2.5 Pro (May 2025) Google DeepMindMay 6, 2025closed142.5
102Qwen3.5-35B-A3B AlibabaFeb 24, 2026open142.5
103GPT-5.5 Instant OpenAIMay 5, 2026closed142.5
104Qwen3-Max AlibabaSep 24, 2025closed142.4
105Claude Haiku 4.5 AnthropicOct 15, 2025closed142.4
106o1 OpenAIDec 17, 2024closed141.9
107Gemma 4 26B A4B Google DeepMindApr 2, 2026open141.8
108Claude Sonnet 4 AnthropicMay 22, 2025closed141.7
109Gemini 2.5 Flash (May 2025) Google DeepMindMay 20, 2025closed141.5
110DeepSeek-R1 (May 2025) DeepSeekMay 28, 2025open141.3
111Mistral Medium 3.5 Mistral AIApr 28, 2026open141.3
112Claude 3.7 Sonnet AnthropicFeb 24, 2025closed141.2
113Gemini 2.5 Flash (Jun 2025) Google DeepMindJun 17, 2025closed140.8
114o3-mini OpenAIJan 31, 2025closed140.3
115Grok-3 mini xAIApr 9, 2025closed140.3
116Kimi K2 (Jul 2025) MoonshotJul 12, 2025open140.1
117Gemini 2.5 Flash (Apr 2025) Google DeepMindApr 17, 2025closed140.0
118gpt-oss-120b OpenAIAug 5, 2025open139.9
119DeepSeek-V3.1 DeepSeekAug 21, 2025open139.9
120Qwen3-30B-A3B-Thinking (Jul 2025) AlibabaJul 30, 2025open139.6
121Qwen3.5-9B AlibabaFeb 24, 2026open139.5
122GPT-5 nano OpenAIAug 7, 2025closed139.4
123Qwen3-235B-A22B AlibabaApr 28, 2025open139.3
124DeepSeek-R1 DeepSeekJan 20, 2025open139.0
125Qwen3-235B-A22B-Instruct (Jul 2025) AlibabaJul 25, 2025open138.9
126Qwen3-32B AlibabaApr 28, 2025open138.5
127Grok 3 xAIApr 9, 2025closed138.3
128Qwen3-14B AlibabaApr 28, 2025open138.2
129gpt-oss-20b OpenAIAug 5, 2025open137.8
130QwQ-32B AlibabaMar 5, 2025open137.6
131DeepSeek-R1-Distill-Qwen-32B DeepSeekJan 20, 2025open137.4
132Qwen3-30B-A3B-Instruct (Jul 2025) AlibabaJul 29, 2025open137.4
133GPT-4.1 OpenAIApr 14, 2025closed136.8
134GPT-4.5 OpenAIFeb 27, 2025closed136.7
135Qwen3-30B-A3B AlibabaApr 28, 2025open136.2
136Qwen3-8B AlibabaApr 28, 2025open136.2
137DeepSeek-V3 (Mar 2025) DeepSeekMar 24, 2025open135.9
138o1-mini OpenAISep 12, 2024closed135.8
139DeepSeek-R1-Distill-Qwen-14B DeepSeekJan 20, 2025open135.4
140Gemini 2.0 Flash Thinking (Jan 2025) Google DeepMind, GoogleJan 21, 2025closed135.4
141Gemini 2.0 Pro Google DeepMindFeb 5, 2025closed135.1
142GPT-4.1 mini OpenAIApr 14, 2025closed135.0
143o1-preview OpenAISep 12, 2024closed134.8
144Gemini 2.0 Flash (Dec 2024) Google DeepMind, GoogleDec 6, 2024closed134.7
145Gemini 2.0 Flash (Feb 2025) Google DeepMind, GoogleFeb 5, 2025closed134.7
146Mistral Medium 3 Mistral AIMay 7, 2025closed134.1
147Gemini 2.5 Flash-Lite (Jun 2025) Google DeepMindJun 17, 2025closed133.9
148Claude 3.5 Sonnet (October 2024) AnthropicOct 22, 2024closed133.6
149Magistral Small 1.0 Mistral AIJun 10, 2025open133.2
150Qwen2.5-Max AlibabaJan 25, 2025closed132.5
151DeepSeek-V3 DeepSeekDec 26, 2024open132.3
152Llama 4 Maverick Meta AIApr 5, 2025open132.2
153Gemini 1.5 Pro (Sept 2024) Google DeepMindSep 24, 2024closed131.7
154Mistral Small 3.2 Mistral AIJun 20, 2025open131.7
155Magistral Small 1.2 Mistral AISep 18, 2025open131.4
156Grok-2 (Dec 2024) xAIDec 12, 2024closed130.5
157Phi-4 Microsoft ResearchDec 12, 2024open130.4
158Claude 3.5 Sonnet AnthropicJun 20, 2024closed130.0
159Gemma 3 27B Google DeepMindMar 11, 2025open130.0
160Llama 4 Scout Meta AIApr 5, 2025open129.6
161GPT-4.1 nano OpenAIApr 14, 2025closed129.6
162Gemini 1.5 Flash (Sep 2024) Google DeepMindAug 27, 2024closed129.4
163GPT-4o (May 2024) OpenAIMay 13, 2024closed129.0
164Qwen2.5-72B AlibabaSep 19, 2024open129.0
165Llama 3.1-405B Meta AIJul 23, 2024open128.8
166GPT-4o (Aug 2024) OpenAIAug 6, 2024closed128.8
167GPT-4o (Nov 2024) OpenAINov 20, 2024closed128.8
168Qwen2.5-32B AlibabaSep 17, 2024open128.5
169Mistral Large 2 (Nov 2024) Mistral AINov 15, 2024open128.5
170Mistral Large 2 (Jul 2024) Mistral AIJul 24, 2024open127.5
171Mistral Small 3.1 Mistral AIMar 17, 2025open127.5
172Llama 3.3 70B Meta AIDec 6, 2024open127.3
173GPT-4 Turbo (Apr 2024) OpenAIApr 9, 2024closed127.2
174Claude 3.5 Haiku AnthropicOct 22, 2024closed127.2
175Mistral Small 3 Mistral AIJan 25, 2025open127.1
176Claude 3 Opus AnthropicFeb 29, 2024closed126.9
177Gemini 1.5 Pro (May 2024) Google DeepMindMay 14, 2024closed126.9
178GPT-4o mini OpenAIJul 18, 2024closed126.6
179GPT-4 Turbo (Nov 2023) OpenAINov 6, 2023closed126.5
180GPT-4 (Mar 2023) OpenAIMar 14, 2023closed125.9
181Llama 3.1-70B Meta AIJul 23, 2024open125.9
182Llama 3.2 90B Meta AISep 24, 2024open125.5
183Qwen2-72B AlibabaJun 7, 2024open125.3
184DeepSeek-V2 (MoE-236B, May 2024) DeepSeekMay 7, 2024open124.8
185Amazon Nova Pro AmazonDec 3, 2024closed123.8
186Gemma 3 12B Google DeepMindMar 12, 2025open123.5
187GPT-4 (Jun 2023) OpenAIJun 13, 2023closed123.1
188Llama 3-70B Meta AIApr 18, 2024open122.9
189Gemini 1.5 Flash (May 2024) Google DeepMindMay 14, 2024closed122.6
190Gemma 2 27B Google DeepMindJun 24, 2024open122.1
191Mistral Large Mistral AIFeb 26, 2024closed122.0
192Mixtral 8x22B Mistral AIApr 17, 2024open122.0
193phi-3-small 7.4B MicrosoftApr 23, 2024open121.8
194phi-3-medium 14B MicrosoftApr 23, 2024open121.2
195Claude 3 Sonnet AnthropicFeb 29, 2024closed120.7
196Claude Instant AnthropicAug 9, 2023closed120.3
197Claude 2 AnthropicJul 11, 2023closed120.1
198Gemma 2 9B Google DeepMindJun 24, 2024open119.8
199Qwen2.5-Coder-32B AlibabaSep 18, 2024open119.5
200Claude 2.1 AnthropicNov 21, 2023closed119.3
201Command R+ Cohere, Cohere Labs (formerly Cohere for AI)Aug 30, 2024open119.3
202Mistral NeMo Mistral AIJul 18, 2024open118.7
203GPT-3.5 Turbo (Nov 2023) OpenAINov 6, 2023closed118.5
204Mixtral 8x7B Mistral AIDec 11, 2023open118.5
205Qwen2.5-7B AlibabaSep 19, 2024open118.5
206Claude 3 Haiku AnthropicMar 7, 2024closed118.3
207Ministral 3B Mistral AIOct 16, 2024closed118.1
208Yi-34B 01.AINov 2, 2023open117.4
209phi-3-mini 3.8B MicrosoftApr 23, 2024open117.4
210Stable Beluga 2 Stability AIJul 20, 2023open117.1
211Gemini 1.0 Pro Google DeepMindDec 13, 2023closed117.0
212Llama 3.1-8B Meta AIJul 23, 2024open116.6
213Llama 3-8B Meta AIApr 18, 2024open116.5
214Gemma 3 4B Google DeepMindMar 12, 2025open116.0
215GPT-3.5 Turbo (Jan 2024) OpenAIJan 25, 2024closed115.7
216Llama 2-70B Meta AIJul 18, 2023open113.8
217GPT-3.5 Turbo (Jun 2023) OpenAIJun 13, 2023closed113.4
218Qwen2.5-Coder (7B) AlibabaSep 18, 2024open113.1
219Qwen-14B AlibabaSep 24, 2023open113.0
220Mistral 7B v0.1 Mistral AISep 27, 2023open112.2
221Falcon-180B Technology Innovation InstituteSep 6, 2023open112.1
222Gemma 7B Google DeepMindFeb 21, 2024open112.0
223DeepSeek LLM 67B DeepSeekNov 29, 2023open110.5
224LLaMA-65B Meta AIFeb 24, 2023open110.2
225Falcon 2 11B Technology Innovation InstituteMay 9, 2024open109.6
226Mistral 7B v0.3 Mistral AISep 27, 2023open109.0
227Phi-2 MicrosoftDec 12, 2023open107.9
228Nemotron-4 15B NvidiaFeb 26, 2024closed107.7
229LLaMA-33B Meta AIFeb 24, 2023open107.4
230Qwen-7B AlibabaSep 28, 2023open106.8
231Llama 2-13B Meta AIJul 18, 2023open106.2
232Llama 2-34B Meta AIJul 18, 2023closed105.2
233StarCoder 2 15B Hugging Face, ServiceNow, Nvidia, BigCodeFeb 20, 2024open105.0
234Yi 6B 01.AINov 2, 2023open104.8
235Falcon-40B Technology Innovation InstituteMar 15, 2023open104.5
236Baichuan2-13B BaichuanSep 6, 2023open103.2
237Qwen2.5-Coder (1.5B) AlibabaSep 18, 2024open103.0
238Llama 3.2 1B Meta AISep 24, 2024open102.0
239INTELLECT-1 Prime Intellect, Hugging Face, Arcee AINov 29, 2024–100.9
240MPT-30B MosaicMLJun 22, 2023open100.7
241LLaMA-13B Meta AIFeb 24, 2023open100.6
242Llama 2-7B Meta AIJul 18, 2023open99.1
243LLaMA-7B Meta AIFeb 24, 2023open96.7
244Baichuan 2-7B BaichuanSep 20, 2023open96.3
245DeepSeek Coder 33B DeepSeek, Peking UniversityNov 2, 2023open96.3
246Falcon-7B Technology Innovation InstituteApr 24, 2023open95.1
247MPT-7B MosaicMLMay 5, 2023open94.6
248Gemma 2B Google DeepMindFeb 21, 2024open94.2
249StarCoder 2 7B Hugging Face, ServiceNow, Nvidia, BigCodeFeb 20, 2024open93.6
250XGen-7B SalesforceJun 27, 2023open93.4
251Phi-1.5 MicrosoftSep 11, 2023open91.5
252Baichuan1-7B BaichuanJun 1, 2023open90.5
253Dolly 2.0-12b DatabricksApr 11, 2023open89.7
254DeepSeek Coder 6.7B DeepSeek, Peking UniversityNov 2, 2023open89.6
255StarCoder 2 3B Hugging Face, ServiceNow, Nvidia, BigCodeFeb 22, 2024open88.8
256Cerebras-GPT-13B Cerebras SystemsMar 20, 2023open83.2
257DeepSeek Coder 1.3B DeepSeek, Peking UniversityNov 2, 2023open63.6

The best open-weights models

#ModelReleasedOverall rankECI
1Kimi K3 MoonshotJul 16, 202615157.4
2DeepSeek V4 Pro 0813 DeepSeekAug 13, 202629155.3
3DeepSeek V4.1 Flash DeepSeekSep 9, 202633154.9
4DeepSeek V4 Flash 0731 DeepSeekJul 31, 202636154.5
5GLM-5.3-Flash Z.ai (Zhipu AI)Aug 20, 202647151.9
6GLM-5.2 Z.ai (Zhipu AI)Jun 16, 202649151.8
7Kimi K2.6 MoonshotApr 20, 202650151.1
8Inkling-Small Thinking MachinesJul 15, 202652150.2
9Kimi K2.7 Code MoonshotJun 12, 202655150.0
10GLM-5.1 Z.ai (Zhipu AI)Apr 7, 202656149.8
11Qwen 3.8 27B AlibabaAug 14, 202658149.4
12DeepSeek-V4-Pro DeepSeekApr 24, 202661149.1
13Inkling Thinking MachinesJul 15, 202663148.5
14Kimi K2.5 MoonshotJan 27, 202664148.0
15MiniMax-M3 MiniMaxJun 1, 202669146.9

The best model of each organization

OrganizationBest modelOverall rankECI
AnthropicClaude Opus 5.51167.3
OpenAIGPT-6 Astra2166.4
MoonshotKimi K315157.4
GoogleGemini 3.7 Flash16157.3
MetaMuse Spark 1.319156.8
AlibabaQwen 3.8 Max22156.4
xAIGrok 4.623156.4
Z.ai (Zhipu AI)GLM-5.327155.6
DeepSeekDeepSeek V4 Pro 081329155.3
Thinking MachinesInkling-Small52150.2
MiniMaxMiniMax-M369146.9
NvidiaNemotron 3 Ultra77146.2
Mistral AIMistral Medium 3.5111141.3
MicrosoftPhi-4157130.4
AmazonAmazon Nova Pro185123.8
CohereCommand R+201119.3
01.AIYi-34B208117.4
Stability AIStable Beluga 2210117.1
Technology Innovation InstituteFalcon-180B221112.1
Hugging FaceStarCoder 2 15B233105.0
BaichuanBaichuan2-13B236103.2
Prime IntellectINTELLECT-1239100.9
DatabricksMPT-30B240100.7
SalesforceXGen-7B25093.4
CerebrasCerebras-GPT-13B25683.2

The leaders on each benchmark

Benchmark1.2.3.
METR Time HorizonClaude Mythos Preview (Early) 17 h 25 minClaude Opus 4.6 (unknown thinking) 11 h 59 minGemini 3.1 Pro Preview 6 h 24 min
MMLUGPT-4o 88.1%Claude 3.5 Sonnet (20241022) 87.3%DeepSeek v3 87.2%
GPQA DiamondGPT-6 Astra (max) 95.8%Claude Sonnet 5.5 (max) 95.6%Gemini 3.8 Flash (high) 95.4%
Humanity's Last ExamGPT 6 Astra 54.8%Fable 5.1 (xhigh) 46.5%Gemini 3.1 Pro Preview (thinking high) 46.4%
SimpleQA VerifiedGPT-6 Astra (max) 75.6%GPT-6.1 Sol (max) 73.9%Gemini 3.1 Pro Preview 73.5%
GSM8KDS-Coder-V2-Instruct 94.5%Qwen2.5-Coder-14B-Instruct 94.2%Qwen2.5-Coder-32B-Instruct 93.0%
MATH Level 5GPT-5 (high) 98.1%o4-mini (high) 97.8%o3 (high) 97.8%
FrontierMath (Tiers 1–3)GPT-6 Astra (max) 93.7%GPT-6.1 Sol (max) 93.7%Claude Opus 5.5 (max) 91.2%
SWE-bench VerifiedClaude Opus 4.7 (max) 83.5%GPT-5.5 (xhigh) 80.6%Gemini 3.5 Flash 79.3%
Aider PolyglotGPT-5 (high) 88.0%o3-pro 84.9%Gemini 2.5 Pro Preview (Jun 2025) 83.1%
Terminal-BenchGPT-5.5 84.7%GPT-5.4 81.8%Gemini 3.1 Pro 80.2%
ARC-AGIClaude Fable 5 (XHigh) 98.5%Gemini 3.8 Flash (High) 98.5%GPT-6 Astra (XHigh) 98.5%
ARC-AGI-2GPT-6 Astra (Max) 95.0%GPT-6.1 Sol (Max) 94.2%Claude Opus 5.5 (High) 93.3%
HellaSwagGPT-4 95.3%LLaMA-3.1 405B Base 89.2%Falcon-180B 89.0%
WinoGrandeLLaMA-3.1 405B Base 89.2%Claude 3 Opus 88.5%GPT-4 87.5%

Source and license

Data: Epoch AI, “Capabilities & benchmarking”, retrieved on October 12, 2026, licensed CC BY 4.0. ABXK built the rankings from this data and changed no value. Logos from the organizations' websites.

How to use the ranking

Start with the question, not with place one. For a product that has to run on your own servers, only the open-weights list counts. For agents that work on their own for hours, the time horizon on the benchmark page says more than the overall index. And for a direct choice between two or three candidates, tick them in the LLM database and read their results side by side.