AI Model Providers

AI21 Labs

2 models

AI21 Labs builds the Jamba family of hybrid SSM-Transformer models with long context windows, offered via its own API with per-token pricing.

Runway

33 models

Runway builds frontier video-generation models (Gen-4, Aleph) plus hosted partner models (Veo, Seedance), billed per second of generated video via its API.

StepFun

10 models

StepFun (阶跃星辰) is a Chinese AI lab building the Step series of multimodal foundation models (text, vision and audio), offered via its own API with per-token pricing and prompt caching.

Cloudflare Workers AI

47 models

Cloudflare Workers AI runs open models (Llama, Qwen, Mistral, Gemma, GPT-OSS, plus embeddings and speech) on Cloudflare's global edge network, billed per token in USD (also denominated in Neurons).

CoralBricks

3 models

CoralBricks serves open models (GLM, DeepSeek) quantized to NVFP4 / MXFP4 on its own GPUs behind an OpenAI-compatible API (Chat Completions, Responses and background mode), billed per token in USD from prepaid credits. Cached input is billed at $0 on every model; novel input under prompt-cache retention bills at a separate cache-write rate.

DeepInfra

172 models

DeepInfra is a serverless inference marketplace running open models (Llama, Qwen, DeepSeek, GLM, Kimi and more) on managed GPUs, billed per token with no idle cost — among the cheapest hosted open-model APIs.

Fireworks AI

34 models

Fireworks AI is a high-performance serverless inference platform for open and open-weight frontier models (DeepSeek, GLM, Qwen, Kimi, MiniMax, GPT-OSS), offering Standard and Priority throughput tiers with cached-input discounts.

Novita AI

176 models

Novita AI is an open-model inference aggregator with an OpenAI-compatible API, hosting DeepSeek, Llama, Qwen, GLM, Kimi and MiniMax models with per-token pricing and prompt caching.

Ollama

21 models

Ollama runs open models locally and, through Ollama Cloud, hosts them behind an API (native, OpenAI- and Anthropic-compatible) billed per token in USD, with off-peak rates on some models. Its Pro, Max and Team subscriptions are a monthly fee bundling dollars of usage credits spent at the same per-token rates.

OpenCode

105 models

OpenCode sells access to the same models two ways. Zen is a pay-per-token API gateway for the OpenCode coding agent, carrying GPT, Claude, Gemini and Grok alongside open models (Qwen, DeepSeek, MiniMax, GLM, Kimi) — several of them free for a limited time. OpenCode Go is a $10/month subscription over the open models only, with its allowance published in dollars of usage.

SambaNova Cloud

7 models

SambaNova Cloud runs open models (DeepSeek, Llama, Qwen) on custom RDU hardware for record-fast token throughput, billed per token.

Tencent's coding assistant. CodeBuddy and the WorkBuddy desktop agent bill against one shared credit balance, so a subscription covers both.

GitHub's coding assistant, sold only as a subscription. Billing moved from premium requests to AI credits, published at 1 credit = $0.01 USD.

Ollama

21 models

Ollama runs open models locally and, through Ollama Cloud, hosts them behind an API (native, OpenAI- and Anthropic-compatible) billed per token in USD, with off-peak rates on some models. Its Pro, Max and Team subscriptions are a monthly fee bundling dollars of usage credits spent at the same per-token rates.

OpenCode

105 models

OpenCode sells access to the same models two ways. Zen is a pay-per-token API gateway for the OpenCode coding agent, carrying GPT, Claude, Gemini and Grok alongside open models (Qwen, DeepSeek, MiniMax, GLM, Kimi) — several of them free for a limited time. OpenCode Go is a $10/month subscription over the open models only, with its allowance published in dollars of usage.

Alibaba Cloud's coding assistant, renamed from 通义灵码 (Lingma) on 2026-05-20 and repriced onto a Credits quota the same day.

ByteDance's AI IDE. Paid tiers currently sell queue priority (速通) rather than an allowance; a credit-based plan was announced for 2026-07-30.