AI21 Models

Q: Which AI21 model should I start with?

baidu-deepseek-v3.2-think at $0.274/M input — the cheapest entry here that declares a token price. Move up to jimeng-3.0-pro when answer quality matters more than cost.

Q: How is cached input billed?

The Cache read column is the rate for input tokens served from the prompt cache — for example deepinfra-glm-5 bills cache hits at 20% of the input rate and glm-5.2 bills cache hits at 25% of the input rate. Cache write is the surcharge for putting a prompt into the cache in the first place, and only a few upstreams bill it separately. A dash in either column means the catalog carries no cache rate for that model, so plan on paying the full input rate.

29 modelsGeneral models from $0.274/M inputUp to 1.05M context

Usage

Last 2 days · 2026-09-04 to 2026-09-11

Tokens

6.2K

Requests

Models in use

2 of 29

Tokens per day, stacked by model

Which models that traffic went to

Share of 6.2K tokens. 1 models with traffic report no token counts and cannot be ranked here, including glm-image — they are in the request view.

The two views disagree on purpose: a model can take a large share of the calls and a small share of the tokens — many short requests — or the reverse. Which one matters depends on whether your cost is driven by call volume or by prompt length. Measured on AIHubMix over the last 2 days, counting the 29 model IDs listed on this page; traffic routed through upstream-specific IDs that are not in the public catalog is not included.

All 29 AI21 Models

Open in model list

AI21 models on AIHubMix with input and output modalities, context length, maximum output, price per million tokens including cache read and cache write rates, and measured throughput and latency.
	Modalities
gemini-3.1-flash-lite-preview	Takes text, vision, audio, video, PDF. Output modality not published.	1.05M	66K	$2$2/M	—	—	—	—
glm-5.2	Takes text. Output modality not published.	1M	131K	$1.1268$3.9438/M	$0.2817/M	—	107 tok/s	0.75 s
doubao-seed-2-0-lite-260215	Takes text, vision, video. Output modality not published.	256K	128K	$2$2/M	$0.4/M	—	—	—
minimax-m2.5	Takes text. Output modality not published.	205K	—	$0.288$1.152/M	—	—	—	—
deepseek-v3.2-think	Takes text. Output modality not published.	164K	—	$0.274$0.411/M	—	—	15 tok/s	0.36 s
baidu-deepseek-v3.2-think		—	—	$0.274$0.411/M	—	—	—	—
siliconflow-deepseek-v3.2		—	—	$0.274$0.411/M	—	—	4 tok/s	3.03 s
sophnet-deepseek-v3.2		—	—	$0.274$0.411/M	—	—	8 tok/s	1.73 s
mm-minimax-m2.5		—	—	$0.288$1.152/M	—	—	—	—
sophnet-minimax-m2.5		—	—	$0.338$1.352/M	—	—	—	—
baidu-kimi-k2.5		—	—	$0.548$2.877/M	—	—	—	—
s-kimi-k2-thinking		—	—	$0.548$2.192/M	—	—	—	—
sf-kimi-k2-thinking		—	—	$0.548$2.192/M	—	—	—	—
sophnet-kimi-k2.5		—	—	$0.548$2.877/M	—	—	—	—
azure-deepseek-v3.2		—	—	$0.58$1.68/M	—	—	—	—
azure-deepseek-v3.2-speciale		—	—	$0.58$1.68/M	—	—	—	—
azure-kimi-k2-thinking		—	—	$0.6$2.5002/M	—	—	—	—
azure-kimi-k2.5		—	—	$0.6$3/M	—	—	—	—
deepinfra-glm-5		—	—	$0.88$2.816/M	$0.176/M	—	—	—
alicloud-deepseek-v3.2		—	—	$2$2/M	$0.4/M	$2.5/M	—	—
alicloud-glm-5		—	—	$2$2/M	$0.4/M	—	—	—
baidu-glm-5		—	—	$2$2/M	—	—	—	—
glm-image	Takes text, returns vision.	—	—	$2$2/M	—	—	—	—
sophnet-glm-5		—	—	$2$2/M	—	—	—	—
xiaomi-mimo-v2-omni		—	—	$2$2/M	—	—	—	—
xiaomi-mimo-v2-pro		—	—	$2$2/M	—	—	—	—
aihubmix-Jamba-1-5-Large		—	—	$2.2$8.8/M	—	—	—	—
o1-global		—	—	$15$60/M	$7.5/M	—	—	—
jimeng-3.0-pro		—	—	$20$20/M	—	—	—	—

Prices are USD per million tokens; cache read and cache write are the rates for prompt-cache hits and for writing a prompt into the cache. Throughput and latency are measured on AIHubMix — the same figures the model detail page shows — not vendor claims. A dash means the catalog does not publish that field for that model, which is not the same as the model not supporting it.

AI21 on AIHubMix

Which AI21 model should I start with?

baidu-deepseek-v3.2-think at $0.274/M input — the cheapest entry here that declares a token price. Move up to jimeng-3.0-pro when answer quality matters more than cost.

Why are there several entries for the same model?

Because each row is a route you can call, not a model release. Some IDs name an upstream (azure-, alicloud-, cc-), and some differ only in capitalisation, kept so older integrations keep working.

The catalog does not carry a field saying which of those a given row is, so this page does not sort them into buckets it would have to invent. Every row shows that route’s own price, context and speed — compare those directly, and open a model to see the upstreams that serve it.

How is cached input billed?

The Cache read column is the rate for input tokens served from the prompt cache — for example deepinfra-glm-5 bills cache hits at 20% of the input rate and glm-5.2 bills cache hits at 25% of the input rate. Cache write is the surcharge for putting a prompt into the cache in the first place, and only a few upstreams bill it separately. A dash in either column means the catalog carries no cache rate for that model, so plan on paying the full input rate.

Do I need a separate AI21 account?

No. One AIHubMix key covers every model on this page, and switching between them is a change to the model string — billing, rate limits, and logs stay in one place.