Agnes AI, Muse Code, OpenAI Codex, the curated top picks.
Agnes-3.0 series (Flash, Pro, Pro Beta) served on the Agnes-2.0 platform, usually ~100 TPS (~150 off-peak). Coding-first subscription tiers power mainstream coding tools; image understanding, image and video models included. Yearly billing is 2 months off.
Sign up →Meta's coding CLI with flat-rate monthly subscriptions instead of per-token billing. US reference pricing: Everyday $5, High $15, Power $50 per month with prompt allowances scaling 1x/5x/20x; Muse Spark models also available pay-as-you-go from $0.10/MTok input.
Sign up →
OpenAI's Codex coding agent for CLI, IDE, web, and desktop. Codex is included across ChatGPT plans; Plus adds expanded Codex usage and Pro tiers provide higher usage allowances.
Sign up →11 providers with a Dracon-owned promo code (plus 3 in Favorites above).
GLM coding plans via Z.ai: $18/month Lite, $80/month Pro, and $168/month Max before quarterly/yearly discounts. The free Z.ai chat surface and paid API/coding-plan limits are separate.
AI coding assistant and learning platform. Learn to code with AI-powered interactive lessons and projects.
Alibaba Cloud Model Studio's international Coding Plan provides Qwen, GLM, Kimi, and MiniMax models for interactive coding tools. Singapore-region new users may also receive a separate model-by-model free quota; pay-as-you-go uses a different API key and endpoint.
Alibaba's international Qwen platform (Singapore entity). Token Plan subscription gives full-modality Qwen model access through an OpenAI-compatible API with a dedicated plan API key; pay-as-you-go also available. International counterpart of the China-side Alibaba Coding Plan.
Privacy-focused AI platform offering private access to multiple open-source models. Founded by ex-Instagram/Meta engineers. Preferred by OpenClaw users.
Open-source API gateway for LLMs. Route requests to 180+ models from 60+ providers with one integration. Self-hostable with usage tracking and cost optimization.
ByteDance's AI platform. 500K free tokens per LLM, 2M vision tokens. Hosted coding models with a coding plan tier.
Kilo Code: open-source AI coding assistant with Kilo Pass subscriptions ($19/$49/$199/mo) for monthly AI credit. KiloClaw (separate product): managed OpenClaw hosting for users and organizations, $55/mo with 1-day free trial. Open-source build with no-cost entry tier.
Private, uncensored AI for text and image generation. No account required to try. Decentralized inference with focus on privacy and free speech.
Sign up →
GPU cloud with extremely low pricing for open AI models.
Cloud platform for deploying full-stack AI apps. 750K+ developers.
The full directory: 33 providers without a Dracon-owned promo code.
MiniMax Token Plan starts at $20/month (Plus) and provides M3/M2.7 multimodal coding access, image and speech models, and shared quota across text, image, and speech workloads.
Sign up →
Open-source AI coding agent (VS Code extension + CLI). Connects to provider APIs (Anthropic, OpenAI, etc.); no own API tier. Bring-your-own-key. Free to install and use; you pay the underlying provider.
Sign up →Unified LLM router: one API key, many providers. The free tier gives you 50 req/day of free models with no card required; the free model pool rotates.
Sign up →
Run open-source LLMs locally or in the cloud. Free local inference on your own hardware. Optional cloud plans for larger models with datacenter-grade GPUs.
Sign up →
Official consumer plans from Google: Google AI Plus, Pro, and Ultra tiers with Gemini access, cloud storage, and region-dependent YouTube benefits. The reviewed UK storefront exposes Plus / Pro / Ultra 5x / Ultra 20x; prices vary by country.
Sign up →
StepFun's open platform providing access to the Step family of large language models with reasoning and agentic capabilities.
Sign up →xAI's Grok models for frontier reasoning, vision, voice, and image/video generation through one API. Usage-based pricing, no minimums.
Sign up →
Open-weight frontier models (V3, R1, V4 Pro) at aggressive unit pricing. Hosted at platform.deepseek.com.
Sign up →Smart LLM routing platform that automatically selects the optimal model based on quality, cost, and speed requirements.
Sign up →
Claude consumer plans and Claude API by Anthropic. Claude Code is included in paid plans; API usage is separate pay-as-you-go.
Sign up →
Fastest inference on custom Wafer Scale Engine hardware. Up to 1M tokens/day free.
Decentralized serverless AI compute platform on the Bittensor network offering LLM inference at highly competitive rates.
Sign up →
Edge AI inference via Cloudflare's global network. Low latency worldwide.
Sign up →
Affordable multi-model AI inference hosting service providing extremely low-cost access to large language models.
Sign up →
Affordable GPU cloud for open AI models. Serverless and dedicated instances.
Fastest inference for open-source LLMs and image models. Serverless, on-demand GPU deployments, and fine-tuning on one platform. $1 free starter credits.
Run AI models directly in GitHub via a unified API. Free tier with rate limits, paid usage beyond.
Sign up →
Fast inference on LPU hardware. Best-in-class throughput for open-weight models.
Sign up →
OpenAI's developer API for browser-direct BYOK Chat. Use an API key from the OpenAI platform; ChatGPT login is not part of this connection flow.
Sign up →
Inference Endpoints from Hugging Face. Managed deployment for open models.
Mistral's consumer chat (chat.mistral.ai). Free / Pro / Team tiers with multipliers on messages, search, libraries, and image gen. Mistral Vibe (agentic coding) on all tiers. La Plateforme API is a separate pay-per-token product.
Sign up →
Nous Research API portal. Home of Hermes and Dischat models.
Sign up → Google's free-tier LLM API at aistudio.google.com. ~500 requests/day on hosted open + proprietary models (rate caps vary per model; check the provider's docs for the exact cap on the model you plan to call), no card required. Separate from the paid Google AI Plus/Pro/Ultra consumer plans.
Sign up →
Optimized inference containers for NVIDIA GPUs. Enterprise-grade performance.
European cloud AI endpoints. Competitive pricing on open models.
API access to Perplexity's Sonar models. Excellent for research and factuality.
Sign up →
Cloud for running open-source models. Easy API for hosted open-source language and image models.
AI cloud: model inference APIs + GPU instances. Open-source models, pay-per-use.
Sign up →
Dataflow architecture optimized for AI inference. High throughput for enterprise.
European cloud provider with generative AI endpoints. 1M free tokens.
Chinese AI API hub with competitive pricing. Good for Qwen and DeepSeek models.
OpenCode Zen: curated open coding models with limited-time free models plus the $10/month Go plan. Built for coding agents with a rotating model lineup.
Sign up →Moonshot AI's Kimi platform offering large language models with long context windows and tool calling support.
Dracon may earn a commission when you sign up through these links. It never changes what you get.