Standard Compute

Which API should you use with DeepSeek Harness?

For DeepSeek Harness with more than DeepSeek models, the best API is a cost-optimized custom model API: in Settings → Models → Add model provider → Custom model API, pick OpenAI Chat Completions, enter https://api.stdcmpt.com/v1 and the model standardcompute, and Standard Compute's router sends each step to the cheapest model that can do it (Claude Fable 5.1 or GPT-6 Astra only when it counts), on a plan from $19 a month with a $20 compute budget. dsh's own route is a DeepSeek account or API key billed per token, which is hard to beat if DeepSeek models are all you need. OpenRouter is pay-per-use too. dsh itself is free and MIT-licensed, so the model provider is the whole bill.

Guide updated 2026-10-15. Check each provider's current terms below.
Get your API key$0.25 trial for eligible new accounts, no card. Plans from $19/mo.
Light or occasional use
If you only run DeepSeek models, a DeepSeek API key billed per token will likely cost the least, especially with cache hits on long sessions. Check DeepSeek's pricing page against a week of your own usage first.
Daily use, cost-optimized
When dsh runs long coding, research or background tasks every day and you want Claude or GPT on the hard steps, a Standard Compute provider sends each step to the cheapest model that can do it, and for every model we vet providers for price, uptime and speed, with automatic fallback. Opt-in Europe or US Compliance rules apply per key, team or company, and we don't train on your prompts or store them. The bill stays at the monthly plan price: requests stop with HTTP 402 when the compute budget is used; running requests finish and nothing is billed on top.

API and subscription options for dsh

OptionPriceWhat you get
DeepSeek APIPay-as-you-goDeepSeek's own models billed per token with a DeepSeek platform key, the default route in DeepSeek Harness, Codewhale and Reasonix. Usually the cheapest route if DeepSeek models are all you use; check the current per-token and cache-hit prices. Source: api-docs.deepseek.com
Standard Compute (us)From $19/moSmart routing sends each request to the cheapest model that can do it, or pin any model by its exact ID. Standard Compute starts at $19/month with a $20 monthly compute budget. Requests stop when that budget is exhausted until renewal or a plan change; there are no automatic overage charges. Requests are not slowed as the budget is used. It is a fixed bill with a finite allowance.
OpenRouterPay-as-you-goChoose models and providers through one API. Usage-based pricing can suit light or bursty work; estimate long agent sessions and review current fees and limits. Source: openrouter.ai
Local model (Ollama / vLLM)Hardware + electricityNo hosted per-token bill, but hardware, context capacity, throughput and maintenance still limit what you can run. Test task quality on your own hardware. Source: docs.ollama.com

Prices and allowances can change. dsh's custom model API fields, protocols and config file come from DeepSeek Harness's model configuration guide and home-path docs, read on October 15, 2026; DeepSeek's prices are linked to its pricing page. DeepSeek Harness: configure models · DeepSeek Harness repo · DeepSeek API pricing. Spot an error? Tell us.

Add Standard Compute as a custom model API in dsh

dsh Settings → Models → Add model provider → Custom model API
Provider ID    standardcompute          (permanent; lowercase)
Base URL       https://api.stdcmpt.com/v1
API protocol   OpenAI Chat Completions
API key        sc_live_…                (your Standard Compute key)
Models         standardcompute
               anthropic/claude-sonnet-5.5
               deepseek/deepseek-v4-pro

Fetch available models lists what the gateway serves, or type the IDs by hand. Pick the model in the session's model picker; it becomes the default for new sessions. The page writes the same settings to ~/.dsh/profiles/web/cordis.patch.yml.

dsh questions

Which API protocol should I choose for a gateway in DeepSeek Harness?
The one the gateway speaks. The Custom model API card offers OpenAI Chat Completions, OpenAI Responses and Anthropic Messages, saved as openai-completions, openai-responses and anthropic-messages, and one provider speaks one protocol, so a gateway that serves two needs two providers. Standard Compute serves Chat Completions at https://api.stdcmpt.com/v1, which is the simplest choice and the one this guide uses.
Can I configure a DeepSeek Harness provider in cordis.patch.yml instead of the Models page?
Yes. The Models page writes ~/.dsh/profiles/web/cordis.patch.yml (the web profile under $DSH_HOME, default ~/.dsh). Under the llm-pi-ai entry's config, add providers.standardcompute with apiKeyEnv: STANDARDCOMPUTE_API_KEY, api: openai-completions, baseURL: https://api.stdcmpt.com/v1 and a models list such as - id: standardcompute. A Cordis override replaces the complete entry config, so keep any other providers in the same entry. Adapters re-read the file on the next request; no restart needed.
Why is there no Effort menu for my custom model in dsh?
A model you enter by hand declares no reasoning levels, so dsh hides the Effort entry and the endpoint's default decides whether the model thinks. Declare levels with reasoningEfforts on that model in cordis.patch.yml if you want the menu back. With Standard Compute's routed standardcompute model, the router picks the model per step.

Try one task before choosing a plan

  1. Check the API format and model configuration supported by your installed dsh version.
  2. Run a small task on a clean branch and review the result, elapsed time and compute used.
  3. Estimate your actual monthly workload before selecting a larger allowance.
Get your API keydsh setup guide →Compare compute budgets →

Record your own results with the workload CSV and report template. These are blank templates, not published benchmark results.

Disclosure: this guide is maintained by Standard Compute, one of the options above. Choose based on task quality, supported features and total cost for your workload. Light usage may cost less on a usage-based API; sustained workloads need enough allowance on any subscription.

All flat-rate plans compared →The cheapest-tokens math →Plan compute for your team →