🤖 Agent League

#	Model	AIM	Năng lực
1	GPT-5.5 Pro OpenAI	94.7	100	NEW
2	Claude Opus 4.8 Anthropic	94.7	100	NEW
3	Nemotron-ToolOrchestra NVIDIA	94.7	100	NEW
4	claude-3-5-sonnet-20241022 Anthropic	94.7	100	NEW
5	Step-3.5-Flash StepFun	94.7	100	NEW
6	Claude Sonnet 4.5 Anthropic	94.7	100	NEW
7	Claude Sonnet 4.5 (HAL Generalist) Anthropic	94.7	100	NEW
8	GLM-4.7 Zhipu AI	91.9	97	NEW
9	Claude Sonnet 4.5 (High) Anthropic	75.8	78	NEW
10	Claude Sonnet 4.5 High (HAL Generalist) Anthropic	73.9	76	NEW
11	GLM-4.7-Flash Zhipu AI	64.7	66	NEW
12	Claude Opus 4.1 (High) Anthropic	63.6	64	NEW
13	Claude Opus 4.1 High (HAL Generalist) Anthropic	60.0	60	NEW
14	Claude Opus 4.6 Anthropic	51.7	50	NEW
15	GPT-5.2 Thinking OpenAI	51.3	50	NEW
16	Claude Opus 4 (High) Anthropic	45.2	43	NEW
17	claude-3-5-sonnet-20240620 Anthropic	43.7	41	NEW
18	Claude Opus 4.1 Anthropic	41.6	39	NEW
19	Claude Opus 4 High (HAL Generalist) Anthropic	39.2	36	NEW
20	Claude Opus 4.5 Anthropic	38.5	35	NEW
21	GPT-5 (medium) OpenAI	34.5	31	NEW
22	MiniMax M2 MiniMax	32.7	29	NEW
23	GPT-4o OpenAI	32.0	28	NEW
24	GPT-5 Medium (HF Open Deep Research) OpenAI	27.5	23	NEW
25	o4-mini (Low) OpenAI	11.1	4	NEW
26	GPT-5.4 OpenAI	8.7	1	NEW
27	GPT-5.2 OpenAI	8.7	1	NEW
28	GPT-5.5 OpenAI	8.7	1	NEW
29	Claude Sonnet 4.6 Anthropic	8.7	1	NEW
30	GPT-5 OpenAI	8.6	1	NEW
31	GPT-5.1 OpenAI	8.6	1	NEW
32	GPT-5.1 Thinking OpenAI	8.6	1	NEW
33	GPT-5.4 mini OpenAI	8.6	1	NEW
34	GPT-5.4 nano OpenAI	8.6	1	NEW
35	MiniMax M2.1 MiniMax	8.6	1	NEW
36	Claude Haiku 4.5 Anthropic	8.5	1	NEW
37	Nova 2 Lite Amazon	8.5	1	NEW
38	Kimi K2 Instruct Moonshot AI	8.4	0	NEW
39	Nemotron 3 Super (120B A12B) NVIDIA	8.4	0	NEW
40	Qwen3-235B-A22B-Thinking-2507 Alibaba Cloud	8.2	0	NEW
41	o3 OpenAI	8.2	0	NEW
42	GPT-5.4 Pro OpenAI	8.0	0	NEW
43	Claude Opus 4.7 Anthropic	8.0	0	NEW
44	gpt-4o-mini OpenAI	8.0	0	NEW
45	Claude Opus 4 Anthropic	8.0	0	NEW
46	GPT-5 Medium (HAL Generalist) OpenAI	8.0	0	NEW

Cách tính điểm — AIM Index

● Năng lực 65%● Phổ biến (mạng xã hội) 25%● Độ mới 10%

GAIAτ-benchWebArena

Nguồn dữ liệu thật (5)

leaderboard.steel.dev ↗github.com ↗llm-stats.com ↗awesomeagents.ai ↗hal.cs.princeton.edu ↗

Điểm minh hoạ (seed). Pipeline R3/R4 sẽ thay bằng AIM Index thật từ benchmark + mạng xã hội.