Pricing

Core models

USDper 1M tokens

Model	Window	Input	Cached	Output
Kimi K3 `moonshotai/Kimi-K3`	Default (ASAP)	2.50	0.25	12.50
	Balanced	2.00	0.20	10.00
	Flex	1.25	0.15	6.25
GLM-5.3 `zai-org/GLM-5.3`	Default (ASAP)	0.98	0.18	3.08
	Balanced	0.50	0.12	2.50
	Flex	0.40	0.08	1.80
GLM-5.3-Flash `zai-org/GLM-5.3-Flash`	Default (ASAP)	0.11	0.02	0.35
	Balanced	0.08	0.02	0.28
	Flex	0.05	0.01	0.18
DeepSeek V4 Pro 0813 `deepseek-ai/DeepSeek-V4-Pro-0813`	Default (ASAP)	0.92	0.04	2.77
	Balanced	0.74	0.03	2.22
	Flex	0.46	0.02	1.39
DeepSeek V4 Flash 0731 `deepseek-ai/DeepSeek-V4-Flash-0731`	Default (ASAP)	0.09	0.02	0.18
	Balanced	0.07	0.02	0.14
	Flex	0.05	0.01	0.09
Kimi-K2.6 `moonshotai/Kimi-K2.6`	Default (ASAP)	1.00	0.20	4.00
	Balanced	0.45	0.20	3.00
	Flex	0.35	0.10	2.00
Gemma 4 31B IT `google/gemma-4-31B-it`	Default (ASAP)	0.40	0.20	0.60
	Balanced	0.12	0.08	0.60
	Flex	0.06	0.02	0.30
Gemma 4 31B IT (NVFP4) `nvidia/Gemma-4-31B-IT-NVFP4`	Default (ASAP)	0.14	0.07	0.40
	Balanced	0.11	0.06	0.32
	Flex	0.07	0.04	0.20
Gemma 4 12B IT `google/gemma-4-12B-it`	Default (ASAP)	0.30	0.15	2.00
	Balanced	0.10	0.07	2.00
	Flex	0.05	0.02	1.00
gpt-oss-120b `openai/gpt-oss-120b`	Default (ASAP)	0.06	0.03	0.40

Models served with the flex completion window exclusively.

USDper 1M tokens

Model	Window	Input	Cached	Output
Qwen3.6 35B A3B `Qwen/Qwen3.6-35B-A3B`	Flex	0.05	0.02	0.40

See Completion Windows for how to use balanced and flex for lower token prices.
- Not all core models support all windows yet. We regularly bring up new models and expand completion window support for existing ones based on demand. If you have a need that’s not represented above, get in touch.
Prompt caching is implicit, based on prefix matching. Optionally, you may use prompt_cache_key as a routing hint to help maximize cache hit rates.
See Models for capabilities and other details on supported models.
To see what these rates add up to on a full agent workload, use the agent cost calculator.