toolathlon

21 models evaluated

#	Model	Provider	Type	Score
1	Claude Opus 4.8	Anthropic	Closed	59.9%
2	Gemini 3.5 Flash	Google	Closed	56.5%
3	GPT-5.5	OpenAI	Closed	55.6%
4	GPT-5.4	OpenAI	Closed	54.6%
5	DeepSeek V4 Pro (Max)	DeepSeek	Open	51.8%
6	Kimi K2.6	Moonshot AI	Open	50%
7	Step 3.7 Flash	StepFun	Open	49.5%
8	DeepSeek V4 Pro (High)	DeepSeek	Open	49%
9	DeepSeek V4 Flash (Max)	DeepSeek	Open	47.8%
10	DeepSeek V4 Pro	DeepSeek	Open	46.3%
11	MiniMax M2.7	MiniMax	Open	46.3%
12	Claude Opus 4.5	Anthropic	Closed	43.5%
13	DeepSeek V4 Flash (High)	DeepSeek	Open	43.5%
14	GPT-5.4 mini	OpenAI	Closed	42.9%
15	DeepSeek V4 Flash	DeepSeek	Open	40.7%
16	Qwen3.6 Plus	Alibaba	Closed	39.8%
17	GLM-5	Z.AI	Open	38%
18	Qwen3.5 397B	Alibaba	Open	36.3%
19	GPT-5.4 nano	OpenAI	Closed	35.5%
20	Kimi K2.5	Moonshot AI	Open	27.8%
21	Qwen3.6-35B-A3B	Alibaba	Open	26.9%