Leaderboard

Research Rankings

Scientific literature and expert-level knowledge work: Humanity’s Last Exam, GPQA Diamond and MMLU-Pro.

Overall767 Coding469 Math90 Reasoning483 Agentic89 Knowledge482 Multimodal452 Coding Agents68 Scientific Reasoning96 Multilingual12 Writing95 Chat & Assistants95 Research99 Tool Calling45 Long Context116 Speed472 Value471 Open Weights183

99 models

#	Model	Podium Score	GPQA Diamond	Humanity’s Last Exam	MMLU-Pro	Speed
01	Kimi K3.1 Highspeed Moonshot AIOpenreasoning	92.0	92%	—	—	130 t/s
02	Seed 2.0 Pro ByteDance	88.9	88.9%	—	—	40 t/s
03	GPT-5.1 OpenAI	88.1	88.1%	—	—	60 t/s
4	Claude Opus 4.5 Anthropic	87.0	87%	—	—	50 t/s
5	ERNIE 5.1 Baidu	87.0	—	—	87%	60 t/s
6	Qwen3.5 122B-A10B AlibabaOpen	86.7	86.6%	—	86.7%	126 t/s
7	GLM-4.5 Air Zhipu AIOpenreasoning	86.0	88%	—	84%	96 t/s
8	GPT-5 OpenAI	85.7	85.7%	—	—	60 t/s
9	Gemma 4 31B GoogleOpen	85.2	—	—	85.2%	35 t/s
10	DeepSeek V3.2 DeepSeekOpen	85.0	—	—	85%	60 t/s
11	Ling 3.1 Flash Ling	84.0	84%	—	—	175 t/s
12	Qwen3.8 Flash Next AlibabaOpen	81.0	81%	—	—	168 t/s
13	Claude Mythos Preview Anthropicreasoningpreview	79.7	94.6%	64.7%	—	80 t/s
14	GLM-5.3 Flash Z.aiOpen	78.5	78.5%	—	—	182 t/s
15	Claude 5.5 Sonnet Anthropicreasoning	78.1	95%	61.2%	—	94 t/s
16	Kimi K3.1 Pro Moonshot AIOpenreasoning	77.8	97.5%	58%	—	45 t/s
17	Gemini 4 Argon Googlereasoningpreview	76.1	94.2%	58%	—	88 t/s
18	GPT-6.1 Sol OpenAIreasoning	75.8	95.5%	56%	—	75 t/s
19	Claude Opus 4.8 Anthropic	75.8	93.6%	57.9%	—	50 t/s
20	Gemini 3 Pro Google	75.8	91.9%	45.8%	89.8%	120 t/s
21	DeepSeek V4 Pro DeepSeekOpen	75.3	90.1%	48.2%	87.5%	66 t/s
22	Kimi K3 Moonshot AIOpenreasoning	74.8	93.5%	56%	—	37 t/s
23	GPT-6 Astra OpenAIreasoning	74.6	96%	53.2%	—	82 t/s
24	Claude Opus 4.7 Anthropic	74.5	94.2%	54.7%	—	80 t/s
25	Qwen3.7 Max Alibaba	74.5	92.4%	41.4%	89.6%	203 t/s
26	Muse Spark Meta	74.0	89.5%	58.4%	—	80 t/s
27	DeepSeek V4 Flash DeepSeekOpen	73.1	88.1%	45.1%	86.2%	113 t/s
28	GLM-5.2 Z.aiOpen	73.0	91.2%	54.7%	—	164 t/s
29	Claude Fable 5 Anthropicreasoning	72.9	92.6%	53.3%	—	71 t/s
30	Claude Opus 5 Anthropic	72.9	93.2%	52.6%	—	60 t/s
31	Gemini 3.1 Pro Google	72.9	94.3%	51.4%	—	136 t/s
32	GPT-5.5 OpenAI	72.9	93.6%	52.2%	—	60 t/s
33	Claude Opus 4.6 Anthropic	72.2	91.3%	53.1%	—	50 t/s
34	Qwen3.7 Plus Alibaba	71.2	90.3%	34.7%	88.5%	54 t/s
35	Gemini 3.1 Pro Preview Googlereasoningpreview	71.0	91%	51%	—	105 t/s
36	Simba 3.2 Independentreasoning	71.0	88%	40%	85%	62 t/s
37	GPT-5.6 Sol OpenAI	70.9	94.6%	47.2%	—	72 t/s
38	Grok 4 Heavy xAIpreview	69.6	88.4%	50.7%	—	50 t/s
39	Claude Sonnet 4.6 Anthropic	69.5	89.9%	49%	—	50 t/s
40	Qwen3.6 Plus Alibaba	69.2	90.4%	28.8%	88.5%	55 t/s
41	Mistral Large 4 Mistralreasoning	68.8	89.5%	48%	—	92 t/s
42	Qwen3.5 397B-A17B AlibabaOpen	68.3	88.4%	28.7%	87.8%	71 t/s
43	Qwen3.8 Max Alibaba	68.1	92.6%	43.6%	—	55 t/s
44	Granite 4.2 30B Instruct IBMOpen	68.0	68%	—	—	78 t/s
45	Muse Spark 1.1 Meta	67.5	89.8%	45.1%	—	208 t/s
46	GPT-5.6 Terra OpenAI	67.4	92.9%	41.8%	—	127 t/s
47	GLM-5.1 Z.aiOpen	67.0	67%	—	—	74 t/s
48	Gemini 3 Flash Google	67.0	90.4%	43.5%	—	120 t/s
49	Claude Sonnet 4.6 (max) Anthropic	67.0	67%	—	—	50 t/s
50	Nemotron 3 Ultra 550B NVIDIAOpen	66.8	87%	26.6%	86.8%	142 t/s
51	Grok 4.5 xAI	66.7	93%	40.3%	—	59 t/s
52	Gemini 3.5 Flash Google	66.6	92.2%	41%	—	267 t/s
53	MiMo v2.6 Flash XiaomiOpen	66.5	87%	46%	—	118 t/s
54	Claude 5.5 Haiku Anthropicreasoning	66.3	88%	44.5%	—	145 t/s
55	GPT-5.4 OpenAI	66.3	92.8%	39.8%	—	60 t/s
56	GPT-5.3 Codex OpenAI	65.7	91.5%	39.9%	—	118 t/s
57	Gemini 3.6 Flash Google	65.6	92.8%	38.3%	—	233 t/s
58	MiniMax M3 MiniMax	65.0	92.9%	37.1%	—	93 t/s
59	GPT-5.6 Luna OpenAI	64.8	92.3%	37.2%	—	175 t/s
60	NVIDIA Nemotron 3.5 Super-VL NVIDIAOpen	64.5	86%	43%	—	110 t/s
61	MiniMax-M2.7 MiniMax	64.0	64%	—	—	40 t/s
62	GPT-5.2 OpenAI	63.5	92.4%	34.5%	—	90 t/s
63	Kimi K2.6 Moonshot AIOpen	63.5	90.5%	36.4%	—	40 t/s
64	DeepSeek V4.1 Flash DeepSeekOpenreasoning	62.0	62%	—	—	214 t/s
65	Kimi K2.7 Code Moonshot AIOpen	61.2	89.6%	32.8%	—	39 t/s
66	Hunyuan Hy3 Tencent	61.0	90.4%	31.6%	—	69 t/s
67	MiMo V2.5 Pro XiaomiOpen	60.2	86.6%	33.8%	—	65 t/s
68	Gemini 3.1 Flash Lite Preview Googlepreview	60.0	82%	38%	—	210 t/s
69	Inkling Thinking Machines	58.5	87.2%	29.7%	—	81 t/s
70	Granite 4.2 8B Instruct IBMOpen	54.0	54%	—	—	115 t/s
71	Mistral Medium 3.5 Mistral	51.0	51%	—	—	60 t/s
72	Agnes 3.0 Flash Agnesreasoning	49.0	49%	—	—	235 t/s
73	Claude Sonnet 5 Anthropic	48.9	80%	17.8%	—	96 t/s
74	Claude 4.5 Haiku Anthropic	48.0	48%	—	—	80 t/s
75	Nova 2.0 Pro Preview Amazonpreview	46.0	46%	—	—	40 t/s
76	Agnes 2.5 Pro Beta Agnes	45.0	45%	—	—	65 t/s
77	GPT-oss-120B OpenAIOpen	43.0	43%	—	—	240 t/s
78	Granite 4.2 3B Instruct IBMOpen	42.0	42%	—	—	148 t/s
79	DeepSeek V4 Flash Vision DeepSeek	36.0	36%	—	—	120 t/s
80	G9v3-39A5B AI9StarsOpen	36.0	36%	—	—	60 t/s
81	G9v3-3B AI9StarsOpen	36.0	36%	—	—	60 t/s
82	GLM-5.3 Zhipu AI	36.0	36%	—	—	90 t/s
83	Grok 4.6 (low) xAI	36.0	36%	—	—	59 t/s
84	Grok 4.6 (medium) xAI	36.0	36%	—	—	59 t/s
85	Grok 4.6 (xhigh) xAI	36.0	36%	—	—	62 t/s
86	HyperNova 60B MultiverseOpen	36.0	36%	—	—	367 t/s
87	KAT-Coder-Pro V2 Kwai	36.0	36%	—	—	103 t/s
88	LFM2.5-2.6B Liquid AIOpen	36.0	36%	—	—	60 t/s
89	LFM2.5-8B-A1B Liquid AIOpen	36.0	36%	—	—	338 t/s
90	LFM2.5-VL-1.6B Liquid AIOpen	36.0	36%	—	—	395 t/s
91	LongCat 2.0 LongCatOpen	36.0	36%	—	—	42 t/s
92	Magistral Small 1.2 MistralOpen	36.0	36%	—	—	60 t/s
93	Qwen3.8 27B (xhigh) AlibabaOpen	36.0	36%	—	—	53 t/s
94	Qwen3.8 27B (low) AlibabaOpen	36.0	36%	—	—	64 t/s
95	Qwen3.8 27B (medium) AlibabaOpen	36.0	36%	—	—	62 t/s
96	Qwen3.8 27B (Non-reasoning) AlibabaOpenreasoning	36.0	36%	—	—	62 t/s
97	Solar Pro 3 Upstage	34.0	34%	—	—	40 t/s
98	GPT-oss-20B OpenAIOpen	32.0	32%	—	—	90 t/s
99	K2 Think V2 K2	31.0	31%	—	—	40 t/s

Podium Score is a weighted blend of five leaderboards. See Methodology for details.

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 ...