Open models, served fast,
and measured in public.
One endpoint for open-weight models. Change the base URL in the client you already use and keep everything else. Nothing is stored by default, so on a zero-retention key there is no copy of your traffic to train on or hand over. Every performance number on this page comes from a benchmark run we keep and will send you.
Speed you can check
Most providers publish a number and no method. These are measured end to end through the same public endpoint you would call, with the raw results kept alongside.
Connect in one line
The API is OpenAI-compatible. Point your existing client at our base URL. Streaming, tool calling and usage accounting all behave the way your SDK already expects.
https://api.creeksidecompute.ai/v1
Drop-in replacement for the OpenAI base URL.
Authorization: Bearer
One key per project. Ask for a rotation and the old key keeps working until you say you have switched.
Chat, completions, models
Server-sent events for streaming. Usage counts on the final chunk.
Use it from your tools
Anything that speaks the OpenAI API connects with a base URL, a key and a model id. No key yet? Chat with the model on Poe instead.
Add a provider
In ~/.codex/config.toml:
[model_providers.creekside] name = "Creekside" base_url = "https://api.creeksidecompute.ai/v1" env_key = "CREEKSIDE_API_KEY" wire_api = "chat" model_provider = "creekside" model = "qwen/qwen3.8-27b"
Add a provider
In opencode.json:
{ "provider": { "creekside": {
"npm": "@ai-sdk/openai-compatible",
"name": "Creekside",
"options": {
"baseURL": "https://api.creeksidecompute.ai/v1",
"apiKey": "{env:CREEKSIDE_API_KEY}" },
"models": { "qwen/qwen3.8-27b": {} } } } }Pick "OpenAI compatible"
In the provider settings:
Base URL https://api.creeksidecompute.ai/v1 API key your key Model qwen/qwen3.8-27b
Through a proxy
Claude Code speaks Anthropic's API, not OpenAI's, so put LiteLLM between you and us:
litellm --model openai/qwen/qwen3.8-27b \
--api_base https://api.creeksidecompute.ai/v1
# then, in another shell
ANTHROPIC_BASE_URL=http://localhost:4000 claudeAdd a provider
In ~/.hermes/config.yaml:
providers: creekside: api: https://api.creeksidecompute.ai/v1 key_env: CREEKSIDE_API_KEY model: default: qwen/qwen3.8-27b provider: creekside
Add a provider
In ~/.openclaw/openclaw.json, under models.providers:
"creekside": {
"baseUrl": "https://api.creeksidecompute.ai/v1",
"apiKey": "${CREEKSIDE_API_KEY}",
"api": "openai-completions",
"models": [{ "id": "qwen/qwen3.8-27b",
"contextWindow": 65536 }] }
# then set the primary model to creekside/qwen/qwen3.8-27bNo key needed
Chat with the same model as CreeksideQwen on Poe. Poe bills you; nothing to set up.
Listings in progress
OpenRouter and NanoGPT listings are under way. Until then, a direct key or Poe.
Models
Open weights only. We list a model when we can serve it well, and we publish the precision we run it at, so you always know what you are getting.
| Model | Context | Precision | Input | Output | Cached input |
|---|---|---|---|---|---|
| Qwen3.8 27B | 65,536 | FP4 | $0.15 | $1.80 | $0.035 |
| Gemma 4 26B-A4B soon | — | FP8 | — | — | — |
| gpt-oss 20B soon | — | MXFP4 | — | — | — |
Prices are US dollars per million tokens. Cached input applies automatically when your prompt shares a prefix with a recent one, which is most of the time for agents and chat.
Two tiers, one difference
The only thing that separates them is whether we keep your request content. Nothing is throttled and nothing is degraded. The discount on the retained tier is what we pay you for the data rights.
Zero retention
- Request and response content is never written to disk
- Only usage metadata is kept, for billing
- The default for every new key
- Suitable where a data processing agreement is required
Retained · 25% off
- Content stored under licence, for as long as we need it
- You grant us rights to the stored content, including licensing it
- Opt in per key or per request with a header
- Useful when you want us to debug a bad response with you
What we do with your traffic
Short version: as little as possible, and it is enforced by how the service is built, not promised in prose.
Nothing is stored
Content is held in memory for the length of the request and then dropped. The accounting record carries token counts and a request id, never text.
Kept under licence
Stored for as long as we need it. Any single request can opt out with the retention header, whatever the key says.
Expires in 13 months
Usage records expire automatically. Ask us to delete a key's history and we delete it.
Nothing exists to use
Content is never written to disk, so there is no copy to train on, license, hand over or lose. The guarantee is structural, not a promise about our intentions.
Nothing on disk
The servers that answer your requests keep nothing on disk.
Ask for the raw runs
Every speed number here comes from a benchmark run we keep, including the ones that went badly. Ask and we send it.
Start
Request a key
Email us and a key comes back within a day, with credit to benchmark us against whatever you use now. Usage is invoiced monthly; self-serve keys are coming.
Change the base URL
One line in your client. Keep your prompts, your SDK and your tooling.
Compare and decide
Run your own load. If we are not faster or cheaper for your shape of traffic, we would rather you knew.