The short answer
Tools compared25
- Lowest starting price
- $5.99 · Mistral AI
- Free plans
- 12 of 25
- Prices published
- 21 of 25
- Sales-led only
- 0
Large language models at a glance
Median starting price$22.50/moHalf start between $8 and $575
Free plan48%16% offer a time-limited trial
Buy online96%Sign up and pay without a sales call
Sales only0%Demo or contact sales to buy
How they charge
Every Large language model, side by side
| Tool | Starts at | Usage price | Free plan | How you buy | Checks | Checked |
|---|---|---|---|---|---|---|
 Mistral AI |
$5.99/mo | – | Yes | Both | 4/8 | Oct 29, 2026 |
 Hugging Face |
$9/mo | $0.03/minute | Yes | Both | 2/8 | Oct 29, 2026 |
 Anthropic Claude |
$17/mo yearly | $10/minute | Yes | Both | 2/8 | Oct 29, 2026 |
| OpenAI | $20/mo yearly | – | Yes | Both | 4/8 | Oct 29, 2026 |
 Upstage |
$100/mo yearly | $0.15/minute | Yes | Both | 3/8 | Oct 29, 2026 |
 Cohere |
$2,000/mo | $3/minute | Yes | Both | 2/8 | Oct 29, 2026 |
 AI21 Labs |
– | $0.20/minute | – | Both | 5/8 | Oct 29, 2026 |
 Alibaba Qwen |
– | – | Yes | Self-serve | 3/8 | Oct 29, 2026 |
 Amazon Bedrock |
– | $2/minute | – | Both | 5/8 | Oct 29, 2026 |
 Anyscale |
– | $0.01/minute | – | Both | 2/8 | Oct 29, 2026 |
 Databricks Mosaic AI |
– | – | Yes | Both | 0/8 | Oct 29, 2026 |
 DeepSeek |
– | – | – | – | 1/8 | Oct 29, 2026 |
 Fireworks AI |
– | $0.01/minute | – | Both | 7/8 | Oct 29, 2026 |
 Gemini Enterprise Agent Platform |
– | $0.09/vcpu-hour | Yes | Both | 3/8 | Oct 29, 2026 |
 Google Gemini |
– | $2/minute | Yes | Both | 4/8 | Oct 29, 2026 |
| Meta Llama | – | – | – | Both | 5/8 | Oct 29, 2026 |
 Microsoft Azure OpenAI Service |
– | – | – | Both | 2/8 | Oct 29, 2026 |
 Nebius |
– | $9.50/minute | – | Both | 1/8 | Oct 29, 2026 |
 OpenRouter |
– | $5.50/minute | Yes | Both | 5/8 | Oct 29, 2026 |
 Perplexity API |
– | $0/minute | – | Both | 3/8 | Oct 29, 2026 |
 Replicate |
– | $0/minute | – | Both | 2/8 | Oct 29, 2026 |
 SiliconFlow |
– | – | – | Both | 5/8 | Oct 29, 2026 |
 Snowflake Cortex AI |
– | – | – | Both | 1/8 | Oct 29, 2026 |
 Together AI |
– | $3/minute | – | Both | 5/8 | Oct 29, 2026 |
 xAI Grok |
– | $2/minute | Yes | Both | 5/8 | Oct 29, 2026 |
The tools
Mistral AIMistral AI builds open-weight and proprietary language models and sells them through an API platform (Studio), an assistant and coding agent (Vibe), custom model training (Forge), and hosted or self-hosted deployment.$5.99 /moHugging FaceHugging Face is a platform for hosting and sharing machine learning models, datasets and apps, with paid compute and team plans.$9 /moAnthropic ClaudeAnthropic makes the Claude family of large language models, sold through chat apps, Claude Code and a pay-per-token API.$17 /moOPOpenAIOpenAI sells ChatGPT workspace plans and API access to its GPT models for individuals, teams and enterprises.$20 /moUpstageUpstage sells Solar large language models and document processing APIs, plus a no-code agent builder, for enterprise document workflows.$100 /moCohereCohere sells enterprise AI: the North agent platform, Compass search, and Command, Embed, Rerank and Parse models, deployable privately or via Model Vault.$2,000 /moAI21 LabsAI21 trains specialized, customer-owned language models on a customer's data, and also sells Jamba models through a usage-priced API.$0.20 /minuteAlibaba QwenQwen is a family of AI models offered through a free assistant app and an OpenAI-compatible API platform.QuoteAmazon BedrockAmazon Bedrock is an AWS service for building generative AI apps and agents on foundation models from many providers.$2 /minuteAnyscaleManaged platform that runs Ray workloads (training, batch inference, data processing, serving) on hosted or customer-owned GPUs.$0.01 /minuteDatabricks Mosaic AIDatabricks is a unified data and AI platform for data engineering, warehousing, analytics, AI agents and apps.QuoteDeepSeekDeepSeek offers large language models through a chat site, apps and an OpenAI/Anthropic-compatible API.QuoteFireworks AIFireworks AI runs an inference platform for open models, with fine-tuning and dedicated GPU deployments.$0.01 /minuteGemini Enterprise Agent PlatformGoogle Cloud platform for building, deploying and governing AI agents and apps with Gemini, third-party and open models.$0.09 /vcpu-hourGoogle GeminiGoogle's Gemini Developer API gives developers pay-per-token access to Gemini models for text, image, audio and video input, with a free tier and a paid tier.$2 /minuteMEMeta LlamaMeta's developer platform for its Muse family of AI models, offering an API, a terminal coding tool and open-weight models.QuoteMicrosoft Azure OpenAI ServiceAzure OpenAI in Foundry Models gives access to OpenAI foundational, reasoning and multimodal models on Azure, with agent integration and fine-tuning.QuoteNebiusNebius is an AI cloud that rents NVIDIA GPU instances, storage and managed services, with a separate Token Factory for managed model inference.$9.50 /minuteOpenRouterOpenRouter is an API gateway that gives access to 500+ AI models from 80+ providers through one OpenAI-compatible interface with pay-as-you-go credits.$5.50 /minutePerplexity APIPerplexity's developer platform sells usage-priced APIs for model access (Agent and Orchestrator), web search, embeddings and structured yes/no decisions.$0 /minuteReplicateReplicate is a hosted platform for running, fine-tuning and deploying machine learning models through an API.$0 /minuteSiliconFlowSiliconFlow is an AI inference cloud with one OpenAI-compatible API for open and commercial LLMs plus image, video and audio models.QuoteSnowflake Cortex AISnowflake is a consumption-priced cloud data platform that also offers AI features for building LLM and ML models on customer data.QuoteTogether AITogether AI runs open-source AI models through APIs and offers GPU clusters, fine-tuning and dedicated hosting.$3 /minutexAI GrokSpaceXSI sells the Grok family of AI models through a usage-priced API covering text, code, images, video and voice, plus consumer and business Grok apps.$2 /minuteHow this list is made
Every price is read from the vendor's own pricing page on the date shown and tied to the line it came from. Tools are ordered by what they cost at a typical volume, then by starting price. No one pays to be listed or ranked.