Map AI inference providers
TinySet maps US inference providers on model, price and funding. TinyResearch explains how the market is splitting.
TinySet
Dataset · 50 rowsUS AI inference providers, with business model, models offered, price per million tokens and funding
| Provider name | Business model | Models offered | Price per million tokens |
|---|---|---|---|
| RunPod | free tier + usage-based | Kimi K3, Qwen3 32B, FLUX.1 Schnell | $3.00 input / $15.00 output |
| Baseten | pay-per-token + usage-based enterprise contracts | mammoth-1, elephant-1, joey-1, Beam, Muse Spark Open, GLM-5.3, GLM-5.3-Flash, GLM-5.3 Fast, GLM-5.2, GLM-5.2 Fast, DeepSeek V4 Pro 0813, DeepSeek V4.1 Flash, DeepSeek V4.1 Flash 1017, DeepSeek-V4-Flash-0731, Kimi K3, Kimi K3.1 Pro, NVIDIA Nemotron 3 Ultra, Nemotron 3.5 Super VL, GPT OSS 120B | from $0.01 input (joey-1) |
| Amazon Bedrock | pay-per-token | Anthropic Claude, Meta Llama, Amazon Nova, Amazon Titan, Mistral AI, Cohere Command, DeepSeek, Google Gemma, OpenAI, Qwen, Stability AI, xAI, AI21 Labs, Writer, TwelveLabs | from $0.04 |
| Modal | pay-per-token + subscription | KIMI K3, QWEN 3.8, GLM 5.3 | H100 ~$3.95/hr (per-second billing) |
| Google Vertex AI | pay-per-token + subscription | Gemini 3.1 Pro Preview, Gemini 3.5 Flash, Gemini 3 Flash Preview, Gemini 3.1 Flash-Lite, Gemini 3 Pro Image, Gemini 2.5 Pro, Anthropic Claude Model Family, Gemma | $2 input / $12 output (Gemini 3.1 Pro Preview, <=200K tokens); $0.25 input / $1.5 output (Gemini 3.1 Flash-Lite) |
| Replicate | usage-based enterprise contracts | black-forest-labs/flux-2-pro, google/nano-banana-pro, bytedance/seedream-4, openai/gpt-image-2, anthropic/claude-opus-4.7, xai/grok-imagine-video, alibaba/wan-3, deepseek-ai/deepseek-r1 | $3.75 input / $10.00 output (DeepSeek R1) |
| Beam | subscription + usage-based | ||
| Cerebrium | free tier + usage-based | Qwen, Llama 3.2, GPT-OSS, Stable Diffusion XL |
TinyResearch
ReportHow is the US inference market splitting by business model, and who is differentiated?
The US AI inference market has undergone a structural transformation. What began as a training-centric infrastructure story has become an inference-dominated economy: inference now accounts for roughly 55–67% of AI infrastructure spending [38][5][12], and the managed inference market surpassed the AI training market for the first time by end-2025 [41]. The global market was valued at approximately $113–132 billion in 2025–2026 and is projected to reach $255–422 billion by 2030–2033 [8][5][12].
