LLM model rankings

Two rankings that do not depend on each other. The first counts real usage — what developers building on uttapen actually send to the API. The second is a value score that puts the toman price, the capabilities and the context window on one scale. Usage figures refresh every half hour.

Best value models

Score out of 100, from three things only: 50% cheapness (the toman cost of one 1,000-word request), 30% capabilities (tools, vision, reasoning, json, files) and 20% context size. The weights never change, so the same model scores the same way twice. It is not a quality benchmark.

#	Model	Capabilities	Context	≈ 1,000 words	Score
1	Gemini 2.5 Flash Lite (batch)	vision · tools +4	1,048,576	94	83.8
2	Qwen3.7 Flash	vision · tools +2	1,000,000	60	82.5
3	Muse Spark 1.2 Contributor	vision · tools +4	1,048,576	113	81.8
4	Muse Spark 1.3 Contributor	vision · tools +4	1,048,576	113	81.8
5	GPT-5 Nano (batch)	vision · tools +3	400,000	85	81.5
6	NNex-N2-Mini	vision · tools +2	262,144	47	80.5
7	ILing 3.0 Flash	tools · reasoning +1	262,144	32	78.7
8	GPT-4.1 Nano (batch)	vision · tools +2	1,047,576	94	77.8
9	Gemini 2.5 Flash Lite	vision · tools +4	1,048,576	189	76.2
10	GLM Flash Latest	vision · tools +2	1,310,720	116	76.2
11	GLM 5.3 Flash	vision · tools +2	1,310,720	123	75.6
12	Solar Pro 4	tools · reasoning +1	524,288	57	74.8
13	Qwen3.5-Flash	vision · tools +2	1,000,000	123	74.7
14	DeepSeek V4 Flash Latest	tools · reasoning +1	1,310,720	79	74.4
15	Mistral Nemo	tools · json	131,072	18	74
16	GPT-5 Nano	vision · tools +3	400,000	170	74
17	Qwen3.5-9B	vision · tools +2	262,144	94	73
18	GPT-5.6 Luna Pro (batch)	vision · tools +3	1,050,000	264	72.6
19	GPT-5.6 Luna (batch)	vision · tools +3	1,050,000	264	72.6
20	MiMo-V2.5	vision · tools +3	1,050,000	158	72.2

Usage figures are aggregated and anonymous; no request content is ever stored or published (privacy).