relay.fast and OpenRouter, compared on the same models
Both services give one API key for many models through an OpenAI-compatible API. This page puts their prices for the same models side by side, with the date of each price, and lists the differences that matter beyond price.
Prices of the same models
OpenRouter prices come from the 2026-10-22 01:04 UTC snapshot. relay.fast prices come from the live catalog, , and are the lowest price among the sources that serve the model now. At that check, the relay.fast price was lower on both input and output for 23 of the 23 models in the table.
| Model, relay.fast and OpenRouter IDs | OpenRouter input / output per 1M tokens (2026-10-22) | relay.fast from, input / output per 1M tokens |
|---|---|---|
Claude Opus 5.5claude-opus-5-5 · anthropic/claude-opus-5.5 | $4 / $20 | $0.111 / $0.556 |
Claude Sonnet 5.5claude-sonnet-5-5 · anthropic/claude-sonnet-5.5 | $2 / $10 | $0.238 / $1.19 |
Claude Haiku 5.5claude-haiku-5-5 · anthropic/claude-haiku-5.5 | $0.1 / $0.5 | $0.0223 / $0.112 |
Claude Fable 5.1claude-fable-5-1 · anthropic/claude-fable-5.1 | $10 / $50 | $2.23 / $11.16 |
Claude Opus 4.8claude-opus-4-8 · anthropic/claude-opus-4.8 | $5 / $25 | $0.139 / $0.694 |
GPT-6.1 Solgpt-6.1-sol · openai/gpt-6.1-sol | $2 / $10 | $0.03 / $0.15 |
GPT-6 Solgpt-6-sol · openai/gpt-6-sol | $2 / $10 | $0.03 / $0.15 |
GPT-6 Astragpt-6-astra · openai/gpt-6-astra | $10 / $50 | $0.1 / $0.5 |
GPT-6 Lunagpt-6-luna · openai/gpt-6-luna | $0.1 / $0.5 | $0.008 / $0.04 |
GPT-5.6 Solgpt-5.6-sol · openai/gpt-5.6-sol | $4 / $20 | $0.075 / $0.45 |
GPT-5.6 Terragpt-5.6-terra · openai/gpt-5.6-terra | $2 / $12 | $0.03 / $0.18 |
GPT-5.6 Lunagpt-5.6-luna · openai/gpt-5.6-luna | $0.2 / $1.2 | $0.016 / $0.096 |
GPT-5.5gpt-5.5 · openai/gpt-5.5 | $5 / $30 | $0.075 / $0.45 |
Gemini 4 Argongemini-4-argon · google/gemini-4-argon | $2 / $10 | $0.32 / $1.6 |
Gemini 3.8 Flashgemini-3.8-flash · google/gemini-3.8-flash | $0.825 / $4.125 | $0.06 / $0.3 |
Grok 4.7grok-4.7 · x-ai/grok-4.7 | $2 / $6 | $0.0133 / $0.04 |
DeepSeek V4.1 Flashdeepseek-v4.1-flash · deepseek/deepseek-v4.1-flash | $0.3 / $1.2 | $0.03 / $0.12 |
DeepSeek V4 Prodeepseek-v4-pro · deepseek/deepseek-v4-pro | $0.9553 / $1.9105 | $0.066 / $0.198 |
GLM-5.3glm-5.3 · z-ai/glm-5.3 | $0.079 / $1.4 | $0.035 / $0.11 |
Kimi K3kimi-k3 · moonshotai/kimi-k3 | $0.239 / $11.936 | $0.15 / $0.75 |
Qwen3.8 Maxqwen3.8-max · qwen/qwen3.8-max-0902 | $2 / $6 | $0.0667 / $0.2 |
MiniMax M3minimax-m3 · minimax/minimax-m3 | $0.3 / $1.2 | $0.05 / $0.2 |
Hy3hy3 · tencent/hy3 | $0.132 / $0.528 | $0.0198 / $0.0792 |
How the numbers were collected
- OpenRouter: the public models API
https://openrouter.ai/api/v1/models, read 2026-10-22 01:04 UTC. The table showspricing.promptandpricing.completiontimes 1,000,000. Individual providers on OpenRouter can charge differently, and fees for buying credits are not included. - relay.fast: the live pricing catalog, the lowest price among the sources that serve each model at the time the page is generated.
- Neither side includes taxes or payment fees. Both price lists change, so treat the table as a dated snapshot, not a promise.
Differences beyond price
- Catalog: OpenRouter listed 468 models on 2026-10-22. relay.fast lists fewer models and focuses on GPT, Claude, Gemini, Grok and leading open models.
- Request formats: relay.fast serves OpenAI Chat Completions and Responses, Anthropic Messages and Gemini generateContent on one key. Check the OpenRouter documentation for the formats it supports.
- Sources: relay.fast prices come from sources with their own capacity and limits. A source can change its price or close on any day; the pricing page always shows the current one. OpenRouter lets you choose between inference providers for a model.
- Setup: relay.fast has one-command installers for Claude Code, Codex, Cursor, Cline, Kilo Code, Gemini CLI and OpenCode in its setup docs.
Which one fits
relay.fast fits when the models you need are in its catalog and the price per token matters most, for example for coding agents that send long contexts. OpenRouter fits when you need a model relay.fast does not list or want to pick the inference provider yourself. Many teams use both. Before moving production traffic, run the same task on each and compare successful completions, latency and total cost.
Run your own comparison
- Pick 20 to 50 real prompts from your application and freeze the model versions.
- Send them through both services with the same settings and record input and output tokens, latency, errors and retries.
- Divide the total cost by the number of successful tasks, not by requests.
- Write down the date; repeat after a price or model change.
Frequently asked questions
Is relay.fast cheaper than OpenRouter?
The table shows both prices for the same models with their dates. Prices change on both services, so compare current prices and the total cost of your own workload before deciding.
Is relay.fast affiliated with OpenRouter?
No. They are independent services. OpenRouter is a trademark of its owner.
Can the same code work with both?
For common OpenAI-compatible calls, yes: change the base URL, key and model ID. Model IDs differ: OpenRouter uses names such as anthropic/claude-sonnet-5.5, relay.fast uses claude-sonnet-5-5.