AI Model Providers
The major AI API providers we track, with how many models each offers and their cheapest option by blended cost. Want the full price table across every provider? See LLM prices today or browse all models.
Maker of the GPT model family (GPT-4o, GPT-4o mini, o-series reasoning models) and text-embedding models.
Cheapest: GPT OSS 20B · $0.00400 in / $0.0180 out per 1M
Maker of the Claude family (Opus, Sonnet, Haiku) with strong reasoning, long context and prompt caching.
Cheapest: Claude Haiku 5.5 · $0.1000 in / $0.5000 out per 1M
Gemini models (via AI Studio and Vertex AI) with very large context windows and multimodal support.
Cheapest: gemma-3-4b-it · $0.0400 in / $0.0500 out per 1M
Managed access to many foundation models (Anthropic, Meta, Mistral, Amazon and more) on AWS.
Cheapest: Gemma 3 4B IT (Bedrock) · $0.0400 in / $0.0800 out per 1M
Open-weight Llama models, typically served via hosting providers; pricing shown reflects a representative host.
Cheapest: Llama 3.2 1B Instruct · $0.0100 in / $0.0100 out per 1M
European lab offering efficient open and commercial models (Mistral Large, Small, and more).
Cheapest: Ministral 3B · $0.0400 in / $0.0400 out per 1M
Cost-efficient models (DeepSeek-V3, DeepSeek-R1) with aggressive pricing and off-peak discounts.
Cheapest: DeepSeek V4.1 Flash · $0.0300 in / $0.1200 out per 1M
Enterprise-focused models (Command family) and strong embedding/rerank models for RAG.
Cheapest: Command R7B · $0.0375 in / $0.1500 out per 1M
Maker of the Grok family (Grok 4.5, Grok 4.3) with very large context windows, reasoning modes and built-in web/X search.
Cheapest: Grok Code Fast 1 · $0.1900 in / $1.43 out per 1M
Search-augmented Sonar models that answer with live web results and citations; every request adds a per-1K search fee on top of token pricing.
Cheapest: Sonar · $1.00 in / $1.00 out per 1M
Maker of the open-weight Kimi family (Kimi K2, K3) with very long context and strong agentic/coding performance; prices shown reflect a representative host where no first-party API rate is published.
Cheapest: Kimi K2 Instruct · $0.4000 in / $1.80 out per 1M
Maker of the Qwen family (open and commercial), spanning small efficient models to large MoE flagships; served first-party via DashScope/Model Studio and by many hosts.
Cheapest: Qwen3.5 4B · $0.0400 in / $0.0700 out per 1M
Maker of the GLM family (open-weight GLM-5.x and turbo variants) with strong coding and tool-use; prices shown reflect a representative host where no first-party API rate is published.
Cheapest: GLM-4.7-Flash · $0.0600 in / $0.4000 out per 1M
Maker of the MiniMax-M series (open and commercial) tuned for agentic and long-context workloads; prices shown reflect a representative host where no first-party API rate is published.
Cheapest: MiniMax-M2.1 Lightning · $0.1200 in / $0.4800 out per 1M
Maker of the open-weight MiMo family, including omni and pro reasoning variants; prices shown reflect a representative host where no first-party API rate is published.
Cheapest: MiMo-V2.5 · $0.1400 in / $0.2800 out per 1M
Maker of the open Nemotron family; served via NVIDIA NIM and many hosts. Prices shown reflect a representative host where no first-party API rate is published.
Cheapest: Nemotron 3.5 Lightning 30B · $0.0250 in / $0.1200 out per 1M
Maker of the MAI in-house model family (including code-focused variants), served via Azure AI Foundry.
Cheapest: Phi 4 · $0.0700 in / $0.1400 out per 1M
Maker of the open-weight Step family of multimodal and flash reasoning models; prices shown reflect a representative host where no first-party API rate is published.
Cheapest: Step 3.5 Flash · $0.1000 in / $0.3000 out per 1M
Maker of the Hunyuan (Hy) family of open and commercial models; prices shown reflect a representative host where no first-party API rate is published.
Cheapest: Hy-MT2-1.8B · $0.0440 in / $0.1770 out per 1M