Ranking

Ranking: Raciocínio

Raciocínio profundo e ciência: GPQA Diamond, HLE e ARC-AGI-2.

Geral767 Código469 Matemática90 Raciocínio483 Agentes89 Conhecimento482 Multimodal452 Agentes de código68 Raciocínio científico96 Multilíngue12 Escrita95 Chat e assistentes95 Pesquisa99 Chamada de ferramentas45 Contexto longo116 Velocidade472 Custo-benefício471 Pesos abertos183

Modelos: 483

#	Modelo	Podium Score	GPQA Diamond	Humanity’s Last Exam	ARC-AGI-2	Velocidade
01	Claude Mythos Preview Anthropicraciocínioprévia	98.8	94.6%	64.7%	—	80 t/s
02	Kimi K3.1 Pro Moonshot AIAbertoraciocínio	98.0	97.5%	58%	—	45 t/s
03	GPT-6 Astra OpenAIraciocínio	96.0	96%	53.2%	—	82 t/s
4	GPT-6.1 Sol OpenAIraciocínio	95.5	95.5%	56%	—	75 t/s
5	Claude 5.5 Sonnet Anthropicraciocínio	95.0	95%	61.2%	—	94 t/s
6	Gemini 4 Argon Googleraciocínioprévia	94.2	94.2%	58%	—	88 t/s
7	Kimi K3.1 Highspeed Moonshot AIAbertoraciocínio	92.0	92%	—	—	130 t/s
8	Claude Fable 5.1 Anthropicraciocínio	92.0	—	—	—	85 t/s
9	Gemini 3.1 Pro Preview Googleraciocínioprévia	91.0	91%	51%	—	105 t/s
10	GLM-4.5 Air Zhipu AIAbertoraciocínio	90.0	88%	—	—	96 t/s
11	GPT-6 Astra (xhigh) OpenAIraciocínio	90.0	—	—	—	76 t/s
12	GPT-6 Astra (high) OpenAIraciocínio	90.0	—	—	—	90 t/s
13	GPT-6 Astra (medium) OpenAIraciocínio	90.0	—	—	—	98 t/s
14	GPT-6 Astra (low) OpenAIraciocínio	90.0	—	—	—	110 t/s
15	Mistral Large 4 Mistralraciocínio	89.5	89.5%	48%	—	92 t/s
16	Simba 3.2 Independentraciocínio	88.0	88%	40%	—	62 t/s
17	Claude 5.5 Haiku Anthropicraciocínio	88.0	88%	44.5%	—	145 t/s
18	GPT-6 Luna OpenAIraciocínio	88.0	—	—	—	120 t/s
19	GPT-5.6 Sol OpenAI	87.6	94.6%	47.2%	—	72 t/s
20	MiMo v2.6 Flash XiaomiAberto	87.0	87%	46%	—	118 t/s
21	Kimi K3 Moonshot AIAbertoraciocínio	86.5	93.5%	56%	—	37 t/s
22	NVIDIA Nemotron 3.5 Super-VL NVIDIAAberto	86.0	86%	43%	—	110 t/s
23	Claude Opus 5.5 Anthropicraciocínio	86.0	—	—	—	68 t/s
24	Claude Fable 5.1 (XHigh) Anthropicraciocínio	85.0	—	—	—	78 t/s
25	Ling 3.1 Flash Ling	85.0	84%	—	—	175 t/s
26	Claude Fable 5.1 (High) Anthropicraciocínio	85.0	—	—	—	92 t/s
27	Claude Fable 5.1 (Medium) Anthropicraciocínio	85.0	—	—	—	104 t/s
28	Claude Fable 5.1 (Low) Anthropicraciocínio	85.0	—	—	—	118 t/s
29	Claude Opus 4.8 Anthropic	82.8	93.6%	57.9%	—	50 t/s
30	Claude Opus 5 Anthropic	82.3	93.2%	52.6%	—	60 t/s
31	Gemini 3.1 Flash Lite Preview Googleprévia	82.0	82%	38%	—	210 t/s
32	Claude Fable 5 Anthropicraciocínio	81.0	92.6%	53.3%	—	71 t/s
33	GPT-5.5 OpenAI	80.4	93.6%	52.2%	85%	60 t/s
34	GPT-6 Luna (xhigh) OpenAIraciocínio	80.0	—	—	—	118 t/s
35	GPT-6 Luna (high) OpenAIraciocínio	80.0	—	—	—	128 t/s
36	GPT-6 Luna (medium) OpenAIraciocínio	80.0	—	—	—	138 t/s
37	GPT-6 Luna (low) OpenAIraciocínio	80.0	—	—	—	150 t/s
38	Claude Opus 5.5 (XHigh) Anthropicraciocínio	78.0	—	—	—	62 t/s
39	Claude Opus 5.5 (High) Anthropicraciocínio	78.0	—	—	—	74 t/s
40	Claude Opus 5.5 (Medium) Anthropicraciocínio	78.0	—	—	—	82 t/s
41	Claude Opus 5.5 (Low) Anthropicraciocínio	78.0	—	—	—	92 t/s
42	GPT-6 Sol OpenAIraciocínio	75.0	—	—	—	220 t/s
43	Muse Spark 1.3 Museraciocínio	75.0	—	—	—	75 t/s
44	Gemini 3.1 Pro Google	73.5	94.3%	51.4%	77.1%	136 t/s
45	Claude Opus 4.7 Anthropic	73.3	94.2%	54.7%	—	80 t/s
46	Claude Mythos 5 Anthropicraciocínioprévia	71.0	—	—	—	50 t/s
47	GPT-6 Sol (xhigh) OpenAIraciocínio	70.0	—	—	—	210 t/s
48	GPT-6 Sol (high) OpenAIraciocínio	70.0	—	—	—	226 t/s
49	GPT-6 Sol (medium) OpenAIraciocínio	70.0	—	—	—	238 t/s
50	GPT-6 Sol (low) OpenAIraciocínio	70.0	—	—	—	250 t/s
51	GPT-5.6 Terra OpenAI	69.6	92.9%	41.8%	—	127 t/s
52	Grok 4.7 (high) xAIraciocínio	68.0	—	—	—	80 t/s
53	Gemini 3.5 Flash Google	66.5	92.2%	41%	—	267 t/s
54	Qwen3.8 Max Alibaba	66.3	92.6%	43.6%	—	55 t/s
55	Qwen3.8 Flash Next AlibabaAberto	66.0	81%	—	—	168 t/s
56	Claude Opus 4.6 Anthropic	66.0	91.3%	53.1%	68.8%	50 t/s
57	Gemini 3.6 Flash Google	65.7	92.8%	38.3%	—	233 t/s
58	GLM-5.2 Z.aiAberto	65.6	91.2%	54.7%	—	164 t/s
59	MiniMax M3 MiniMax	64.8	92.9%	37.1%	—	93 t/s
60	GPT-5.3 Codex OpenAI	62.9	91.5%	39.9%	—	118 t/s
61	Muse Spark 1.1 Meta	62.7	89.8%	45.1%	—	208 t/s
62	GLM-5.3 Flash Z.aiAberto	62.5	78.5%	—	—	182 t/s
63	Grok 4.7 xAIraciocínio	62.0	—	—	—	95 t/s
64	Step-5 StepFunraciocínio	62.0	—	—	—	110 t/s
65	GLM-5.1 Z.aiAberto	62.0	67%	—	—	74 t/s
66	Claude Sonnet 4.6 (max) Anthropic	62.0	67%	—	—	50 t/s
67	Qwen3.7 Max Alibaba	61.0	92.4%	41.4%	—	203 t/s
68	Grok 4.5 xAI	60.8	93%	40.3%	—	59 t/s
69	GPT-5.4 OpenAI	60.1	92.8%	39.8%	73.3%	60 t/s
70	MiniMax-M2.7 MiniMax	60.0	64%	—	—	40 t/s
71	DeepSeek V4.1 Flash DeepSeekAbertoraciocínio	58.0	62%	—	—	214 t/s
72	Gemini 3.8 Flash Googleraciocínio	56.0	—	—	—	302 t/s
73	Gemini 3.8 Flash (Medium Reasoning) Googleraciocínio	56.0	—	—	—	315 t/s
74	Gemini 3.8 Flash (Low Reasoning) Googleraciocínio	56.0	—	—	—	340 t/s
75	GPT-5.1 OpenAI	55.5	88.1%	—	—	60 t/s
76	GPT-5.6 Luna OpenAI	55.1	92.3%	37.2%	—	175 t/s
77	GPT-Image 2 OpenAI	55.0	—	—	—	25 t/s
78	GPT-6 Luna (Non-reasoning) OpenAI	55.0	—	—	—	165 t/s
79	Claude Opus 4 7 Non Reasoning Anthropicraciocínio	55.0	—	—	—	50 t/s
80	Claude Sonnet 5 Non Reasoning Anthropicraciocínio	55.0	—	—	—	50 t/s
81	Claude Opus 4 5 Thinking Anthropicraciocínio	55.0	—	—	—	50 t/s
82	Gpt 5 6 Sol Non Reasoning OpenAIraciocínio	55.0	—	—	—	60 t/s
83	Solar Pro4 Upstageraciocínio	55.0	—	—	—	40 t/s
84	Deepseek V3 2 Reasoning DeepSeekAbertoraciocínio	55.0	—	—	—	40 t/s
85	Claude 4 5 Sonnet Thinking Anthropicraciocínio	55.0	—	—	—	50 t/s
86	Glm 5 1 Non Reasoning Z.aiAbertoraciocínio	55.0	—	—	—	40 t/s
87	Gpt 5 5 Non Reasoning OpenAIraciocínio	55.0	—	—	—	60 t/s
88	Kimi K2 6 Non Reasoning Moonshot AIraciocínio	55.0	—	—	—	40 t/s
89	Claude Sonnet 4 6 Non Reasoning Low Effort Anthropicraciocínio	55.0	—	—	—	50 t/s
90	Glm 5 2 Non Reasoning Z.aiAbertoraciocínio	55.0	—	—	—	40 t/s
91	Gpt 5 6 Terra Non Reasoning OpenAIraciocínio	55.0	—	—	—	60 t/s
92	Claude 4 1 Opus Thinking Anthropicraciocínio	55.0	—	—	—	50 t/s
93	Kimi K2 Thinking Moonshot AIraciocínio	55.0	—	—	—	40 t/s
94	O3 Pro OpenAIraciocínio	55.0	—	—	—	60 t/s
95	Glm 5 Non Reasoning Z.aiAbertoraciocínio	55.0	—	—	—	40 t/s
96	Claude 4 5 Haiku Reasoning Anthropicraciocínio	55.0	—	—	—	50 t/s
97	Qwen3 5 397b A17b Non Reasoning AlibabaAbertoraciocínio	55.0	—	—	—	55 t/s
98	Nova 2 0 Pro Reasoning Medium Amazonraciocínio	55.0	—	—	—	40 t/s
99	Deepseek V4 Pro 0424 Non Reasoning DeepSeekAbertoraciocínio	55.0	—	—	—	40 t/s
100	Claude 4 Opus Thinking Anthropicraciocínio	55.0	—	—	—	50 t/s

O Podium Score é uma mistura ponderada de cinco leaderboards. Saiba mais na metodologia.

Model Name

Provider

Podium Score0.0

Context128K

Price (1M out)$0/M

Speed0 t/s

Architecture & Highlights

Model overview and workload profile.

Quickstart cURL

curl https://api.openrouter.ai/api/v1/chat/completions ...