DeepInfra

Inference host · 57 models · 78 priced routes · deepinfra.com

Models on DeepInfra

About DeepInfra

Caching: Cached input prices listed per model (e.g. $0.06 / $0.015 cached)

Across GlyphCompany profile Funding rounds socials 500 Global

Model	In / out per 1M	Vs headline
Qwen3 Embedding 8B us · The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks.	$0.01 / FreeSame vs headline	Same	us · The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks.
Qwen3 Embedding 4B us · The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks.	$0.02 / FreeSame vs headline	Same	us · The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks.
Llama 3.1 8B Instruct fp8 · Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors.	$0.02 / $0.04Same vs headline	Same	fp8 · Meta's latest class of model (Llama 3.1) launched with a variety of sizes & flavors.
Mistral Nemo fp8 · A 12B parameter model with a 128k token context length built by Mistral in collaboration with NVIDIA.	$0.029 / $0.03+39% vs headline	+39%	fp8 · A 12B parameter model with a 128k token context length built by Mistral in collaboration with NVIDIA.
gpt-oss-20b bf16 · gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license.	$0.03 / $0.14+693% vs headline	+693%	bf16 · gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license.
Mistral Small 3 fp8 · Mistral Small 3 is a 24B-parameter language model optimized for low-latency performance across common AI tasks.	$0.05 / $0.08Same vs headline	Same	fp8 · Mistral Small 3 is a 24B-parameter language model optimized for low-latency performance across common AI tasks.
Gemma 3 12B bf16-flex, bf16 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.	$0.04 / $0.12Same vs headline	Same	bf16-flex, bf16 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.
Gemma 3 4B bf16 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.	$0.05 / $0.1+47% vs headline	+47%	bf16 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.
gpt-oss-120b bf16 · 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.	$0.037 / $0.17+319% vs headline	+319%	bf16 · 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.
Gemma 3 12B bf16 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.	$0.05 / $0.15+25% vs headline	+25%	bf16 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.
Nemotron 3.5 Lightning bf16 · NVIDIA Nemotron 3.5 Lightning is an open mixture-of-experts model from NVIDIA, with 3B active parameters out of 30B total.	$0.06 / $0.16+74% vs headline	+74%	bf16 · NVIDIA Nemotron 3.5 Lightning is an open mixture-of-experts model from NVIDIA, with 3B active parameters out of 30B total.
DeepSeek V4 Flash 0731 fp8 · DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total.	$0.06 / $0.18+1% vs headline	+1%	fp8 · DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total.
Gemma 3 27B fp8 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.	$0.08 / $0.16Same vs headline	Same	fp8 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.
Mistral Small 3.2 24B fp8 · Mistral-Small-3.2-24B-Instruct-2506 is an updated 24B parameter model from Mistral optimized for instruction following, repetition reduction, and improved function calling.	$0.075 / $0.2Same vs headline	Same	fp8 · Mistral-Small-3.2-24B-Instruct-2506 is an updated 24B parameter model from Mistral optimized for instruction following, repetition reduction, and improved function calling.
Gemma 3 12B bf16-priority, bf16 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.	$0.075 / $0.225+87% vs headline	+87%	bf16-priority, bf16 · Gemma 3 introduces multimodality, supporting vision-language input and text outputs.
DeepSeek V4 Flash 0423 fp8 · DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window.	$0.09 / $0.18Same vs headline	Same	fp8 · DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window.
Qwen3.5-9B bf16 · Qwen3.5-9B is a multimodal foundation model from the Qwen3.5 family, designed to deliver strong reasoning, coding, and visual understanding in an efficient 9B-parameter architecture.	$0.1 / $0.15+22% vs headline	+22%	bf16 · Qwen3.5-9B is a multimodal foundation model from the Qwen3.5 family, designed to deliver strong reasoning, coding, and visual understanding in an efficient 9B-parameter architecture.
Llama 4 Scout fp8-flex, fp8 · Llama 4 Scout 17B Instruct (16E) is a mixture-of-experts (MoE) language model developed by Meta, activating 17 billion parameters out of a total of 109B.	$0.08 / $0.24Same vs headline	Same	fp8-flex, fp8 · Llama 4 Scout 17B Instruct (16E) is a mixture-of-experts (MoE) language model developed by Meta, activating 17 billion parameters out of a total of 109B.
Qwen3 32B fp8 · Qwen3-32B is a dense 32.8B parameter causal language model from the Qwen3 series, optimized for both complex reasoning and efficient dialogue.	$0.08 / $0.28Same vs headline	Same	fp8 · Qwen3-32B is a dense 32.8B parameter causal language model from the Qwen3 series, optimized for both complex reasoning and efficient dialogue.
Gemma 4 26B A4B fp8 · Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind.	$0.07 / $0.34+62% vs headline	+62%	fp8 · Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind.