Groq

Inference host · 5 models · 5 priced routes · groq.com

Models on Groq

Model	In / out per 1M	Vs headline
Llama 3.1 8B Instruct Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors.	$0.05 / $0.08+130% vs headline	+130%	Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors.
gpt-oss-safeguard-20b gpt-oss-safeguard-20b is a safety reasoning model from OpenAI built upon gpt-oss-20b.	$0.075 / $0.3Same vs headline	Same	gpt-oss-safeguard-20b is a safety reasoning model from OpenAI built upon gpt-oss-20b.
gpt-oss-20b gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license.	$0.075 / $0.3+1710% vs headline	+1710%	gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license.
gpt-oss-120b gpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and general-purpose production use cases.	$0.15 / $0.6+1467% vs headline	+1467%	gpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and general-purpose production use cases.
Llama 3.3 70B Instruct The Meta Llama 3.3 multilingual large language model (LLM) is a pretrained and instruction tuned generative model in 70B (text in/text out).	$0.59 / $0.79+313% vs headline	+313%	The Meta Llama 3.3 multilingual large language model (LLM) is a pretrained and instruction tuned generative model in 70B (text in/text out).

Batch: 50% cost discount compared to synchronous APIs (console.groq.com/docs/batch); does not stack with prompt caching

Caching: There is a 50% discount for cached input tokens (console.groq.com/docs/prompt-caching)