Solo Router logo

Solo Router

A prepaid AI model router that gives you one API key and one balance for 36 open and Chinese chat models such as DeepSeek, Kimi, Qwen, GLM, MiniMax, Llama and Gemma, plus image, video, speech and embedding models, and reports the exact cost of every request in the response.

Last verified: November 3, 2026
Alternative to: OpenRouter
Solo Router app screenshot

Short description

Solo Router is a hosted service that passes your requests on to many different AI models through a single account. Instead of opening separate accounts with DeepSeek, Moonshot (Kimi), Alibaba (Qwen), Z.ai (GLM) and others, you sign in with Google, get one API key (the code your software uses to call the service) and pay from one prepaid balance. To switch models, you change the model name in your request.

It accepts the same request format as OpenAI's chat API, so existing OpenAI-style code, LangChain, Cline, Continue or Open WebUI work once you change the base URL to https://solorouter.ai/api/v1. The catalog has 157 models: 25 regular chat models, 11 "unrestricted" (uncensored) chat models and over 100 image, video, speech, music, embedding and reranking models. You pay per token, a small piece of text, at one listed price per model that already includes Solo Router's margin. There's no subscription. Every response reports its cost in US dollars, and each key can have its own spending cap. Solo Router says it doesn't store prompts or responses, only token counts and cost. It has no Claude or GPT models.

It suits developers who mostly build on open-weight models, meaning models whose files are published for anyone to run, and who want simple prepaid billing. Solo Router launched in late October 2026 and is run by ULA SOFT LTD, a company registered in London, England.

Features

  • One key for many models: A single API key and balance covers 36 chat models from DeepSeek, Z.ai (GLM), Moonshot (Kimi), Alibaba (Qwen), MiniMax, Meta, Google, Mistral, NVIDIA, Xiaomi and Reflection AI, plus image, video and voice models. You switch by changing the model ID, written in full (z-ai/glm-5.3) or short form (glm-5.3).

  • Works with OpenAI-style code: Send requests to https://solorouter.ai/api/v1 with your key. On models that support them, it handles streaming (replies appear word by word), reasoning output, function calling (the model asks your code to run a tool) and JSON mode (replies in a fixed data format).

  • Cost in every response: Each response includes its cost in US dollars in a usage.cost field (speech returns it in an X-Solo-Cost-USD header), and an Activity page lists every request.

  • Spending limits per key: Give each key its own cap. Each request sets aside its maximum possible cost first, and requests stop when the balance or the cap is reached.

  • Long inputs and cheaper repeats: Many chat models have a 1M-token context window, which is how much text the model can take in at once. Repeated prompt openings are billed at a much lower cached-input price, for example $0.008 instead of $0.39 per million tokens on DeepSeek V4.1 Flash.

  • Unrestricted models: 11 uncensored or "abliterated" models (versions with their refusal behavior removed), from $0.091 per million input tokens, aimed at fiction, role-play and security research. The Acceptable Use Policy still bans illegal and sexually explicit content.

  • Image generation and editing: Over 60 image models and edit variants, including FLUX 3, Nano Banana 2.1, Seedream 5.0, Qwen Image 3, Ideogram 4.5 and Recraft V4.1, charged per image or per megapixel. Image links expire after 24 hours.

  • Video in the background: Text-to-video and image-to-video from Kling, Veo, Seedance, Wan, MiniMax, LTX, Luma and more, run as background jobs. You're charged only when the video is ready, and failed videos are free.

  • Speech, music, embeddings and reranking: Qwen3 TTS and ElevenLabs Multilingual v2 voices, Stable Audio 2.5 music, six embedding models such as BGE-M3 and E5 (embeddings turn text into numbers for search) and a BGE reranker that sorts search results by relevance, all on the same key.

  • Browser chat: Sign in and chat with any model from the same balance without writing code.

  • No prompt storage: Solo Router logs only token counts and cost. It says it sends requests only to providers that state they don't train on or keep API inputs and outputs.

Pricing

You top up a prepaid balance and each request is taken out of it. The cost is input tokens times the input price plus output tokens times the output price. Tokens are pieces of words, and Solo Router's guide puts 1,000 tokens at about 750 English words. Input is everything you send (system prompt, chat history and new message), output is what the model writes, and the model's "thinking" counts as output. Chat prices below are in US dollars per million tokens, shown as input / output / cached input.

  • Free sign-up credit: $0.25 when you sign up with Google, one per person. Works on all chat models.

  • Pay as you go: Top up any amount from $5. A 5% card processing fee (minimum $0.50) is added at checkout. No subscription. Credit doesn't expire while the account is active.

  • Markup: Per-token prices include Solo Router's margin, which the vendor says makes them "usually a bit higher" than buying direct. The margin per model isn't published.

  • Refunds: All credit purchases are final and non-refundable. Failed requests aren't charged.

  • DeepSeek V4.1 Flash: $0.39 / $1.56 / $0.008, 1M context.

  • DeepSeek V4 Flash: $0.182 / $0.364 / $0.036, 1M context. The cheapest chat model listed.

  • DeepSeek V4 Pro: $2.262 / $4.524 / $0.452, 1M context. V4 Pro (0813) is $1.716 / $5.148 / $0.057.

  • DeepSeek V3.2: $0.65 / $2.08 / $0.195, 164K context.

  • GLM models: GLM-5.3 $1.82 / $5.72 / $0.338 with 1M context; GLM-5.3 Flash $0.195 / $0.65 / $0.039; GLM-5.2 $1.82 / $5.72 / $0.273.

  • Kimi models: Kimi K3 $3.90 / $19.50 / $0.39 with 262K context; Kimi K2.6 $1.235 / $5.20 / $0.247.

  • Qwen models: Qwen 3.8 Max $2.60 / $7.80 / $0.26 with 1M context; Qwen 3.5 397B A17B $0.715 / $4.55 / $0.144.

  • MiniMax models: MiniMax M3.1 Flash $0.195 / $0.78 / $0.019; MiniMax M2.5 $0.39 / $1.56 / $0.078.

  • Gemma, Llama and Ministral: Gemma 4 31B $0.234 / $0.65; Llama 4 Maverick $0.325 / $1.131; Ministral 3 14B $0.26 / $0.26.

  • NVIDIA, Xiaomi and Reflection: NVIDIA Nemotron models from $0.26 / $0.78; Xiaomi MiMo V2.5 Pro $1.04 / $3.90; Reflection Beam $0.52 / $1.56.

  • Unrestricted models: From $0.091 / $0.52 (GLM 4.7 Flash Heretic) up to $4.875 / $9.75 (Aion 3.5). DeepSeek V4.1 Flash Uncensored is $0.429 / $1.716.

  • Images: From $0.0039 per megapixel (FLUX.1 schnell) to $0.273 per image (Recraft V4.1 Pro). FLUX 3 is $0.062 per image.

  • Video: From $0.052 per second (MiniMax H3 Max Turbo) to $0.65 per second (Seedance 2.5 with audio). Kling 4.0 is $0.204 per second.

  • Speech and music: Qwen3 TTS $0.026 and ElevenLabs Multilingual v2 $0.130 per 1,000 characters. Stable Audio 2.5 is $0.0008 per GPU second.

  • Embeddings and reranking: Embeddings from $0.0117 per million tokens. BGE Reranker v2 M3 is $0.013 per million tokens.

  • Rate limits: Not published for paid use. The launch offer, now ended, was capped at 20 requests a minute.

  • Launch offer (ended): DeepSeek V4.1 Flash Uncensored was free for 48 hours, until November 1, 2026 at 06:58 UTC, with 500K tokens per account per day. The offer page now says it has ended.

  • Enterprise or volume plans: Not published.

Integrations

  • Platforms: Web (browser chat) and an API that developers call from their own code. The research doesn't mention a desktop or mobile app.

  • OpenAI-style SDKs and HTTP clients: Any client where you can set a base URL, including Python requests, JavaScript fetch and curl.

  • LangChain, LlamaIndex and Vercel AI SDK: Point the provider's base URL at Solo Router and use your key.

  • Coding agents: Set Solo Router as a custom provider in Cline, Continue or Cursor and type a model ID.

  • Chat apps: Use Solo Router as the backend for self-hosted chat or role-play apps such as Open WebUI, LibreChat and SillyTavern.

  • n8n: Call models from n8n automation workflows using the same chat request format.

  • Google sign-in: The only sign-in method listed.

Platforms

  • Web
  • API
💡

Frequently asked questions about Solo Router

Is the DeepSeek API free?â–¼

Not on an ongoing basis. Solo Router's launch offer made DeepSeek V4.1 Flash Uncensored free for 48 hours, and it ended on November 1, 2026. New accounts now get $0.25 of free credit when they sign in with Google, and after that you top up from $5. The cheapest DeepSeek model, V4 Flash, costs $0.182 per million input tokens and $0.364 per million output tokens.

How do I get a DeepSeek API key?â–¼

On Solo Router you sign in with Google, open the Keys page and copy the key created for you, or create a new one with its own spending limit. Send requests to https://solorouter.ai/api/v1/chat/completions with a model such as "deepseek/deepseek-v4.1-flash". The same key works for every other model on the service, so you don't need a separate DeepSeek account.

How much does the DeepSeek API cost per million tokens?â–¼

On Solo Router, DeepSeek V4.1 Flash costs $0.39 per million input tokens and $1.56 per million output tokens, V4 Flash costs $0.182 and $0.364, and V4 Pro costs $2.262 and $4.524. Repeated prompt openings are billed at a cached price, as low as $0.008 per million on V4.1 Flash. Solo Router's own example: a chat message with about 2,000 input and 400 output tokens on V4.1 Flash costs about $0.0014, or roughly $1.40 per thousand messages.

What's the best alternative to OpenRouter?â–¼

It depends on which models you need. Solo Router uses the same request format and the same company/model ID style, so switching means changing one base URL. It shows one all-in price per model and puts the cost in every response. It covers 36 open and Chinese chat models but has no Claude or GPT, and Solo Router itself says OpenRouter or the model makers are the better choice if you need those.

How do I use the DeepSeek API in Cline, Cursor or other tools?â–¼

In the tool's "base URL" or "custom provider" setting, enter https://solorouter.ai/api/v1, paste your Solo Router key and type a model ID from the Models page, such as deepseek/deepseek-v4.1-flash. Solo Router lists LangChain, Cline, Continue, Cursor, Open WebUI, LibreChat and SillyTavern as working this way. Streaming and function calling work on models that support them.