From c8382dd8f9e236522578ec4d8ef4616eddb22406 Mon Sep 17 00:00:00 2001 From: Henrik Bakke Dukefoss Date: Mon, 9 Nov 2026 11:42:27 +0100 Subject: [PATCH 1/2] docs(guide): document Standard Compute as an openai_compatible provider example Co-Authored-By: Claude Opus 5.5 --- docs/guide/new-backend.md | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/docs/guide/new-backend.md b/docs/guide/new-backend.md index 07f6a1cc8..992a7f1bb 100644 --- a/docs/guide/new-backend.md +++ b/docs/guide/new-backend.md @@ -29,6 +29,7 @@ A single `base_url` + `api_key` pair lets you point SkillOpt at, for example: | OpenRouter / Fireworks / xAI / … | provider base URL | provider model id | | OrcaRouter | `https://api.orcarouter.ai/v1` | `openai/gpt-5.5` | | Cheaper Inference | `https://api.cheaperinference.com/v1` | `gpt-5.4-mini` | +| Standard Compute | `https://api.stdcmpt.com/v1` | `StandardCompute` | [OrcaRouter](https://www.orcarouter.ai) exposes OpenAI, Anthropic, Google, DeepSeek and other models behind one OpenAI-compatible endpoint. Its model IDs @@ -40,6 +41,12 @@ the bare ID as the model. See its [API docs](https://www.cheaperinference.com/do for the base URL and its [model list](https://www.cheaperinference.com/markets) for the model IDs. +On [Standard Compute](https://standardcompute.com), the `StandardCompute` +model sends each request to the cheapest enabled model that can handle it, +with automatic fallback. To pin a model instead, pass an ID exactly as listed +by `GET https://api.stdcmpt.com/v1/models`. See its +[docs](https://docs.standardcompute.com) for setup. + ### Python API Select and configure the backend directly when embedding SkillOpt as a Python From 13d570384ca16144e82f5c2ea1cf2f0764d4eaf5 Mon Sep 17 00:00:00 2001 From: Henrik Bakke Dukefoss Date: Tue, 17 Nov 2026 17:24:19 +0100 Subject: [PATCH 2/2] docs(guide): pin a Standard Compute model ID for reproducible runs Co-Authored-By: Claude Opus 5.5 --- docs/guide/new-backend.md | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) diff --git a/docs/guide/new-backend.md b/docs/guide/new-backend.md index 992a7f1bb..7eedff2dc 100644 --- a/docs/guide/new-backend.md +++ b/docs/guide/new-backend.md @@ -29,7 +29,7 @@ A single `base_url` + `api_key` pair lets you point SkillOpt at, for example: | OpenRouter / Fireworks / xAI / … | provider base URL | provider model id | | OrcaRouter | `https://api.orcarouter.ai/v1` | `openai/gpt-5.5` | | Cheaper Inference | `https://api.cheaperinference.com/v1` | `gpt-5.4-mini` | -| Standard Compute | `https://api.stdcmpt.com/v1` | `StandardCompute` | +| Standard Compute | `https://api.stdcmpt.com/v1` | `deepseek/deepseek-v4-flash` | [OrcaRouter](https://www.orcarouter.ai) exposes OpenAI, Anthropic, Google, DeepSeek and other models behind one OpenAI-compatible endpoint. Its model IDs @@ -41,11 +41,13 @@ the bare ID as the model. See its [API docs](https://www.cheaperinference.com/do for the base URL and its [model list](https://www.cheaperinference.com/markets) for the model IDs. -On [Standard Compute](https://standardcompute.com), the `StandardCompute` -model sends each request to the cheapest enabled model that can handle it, -with automatic fallback. To pin a model instead, pass an ID exactly as listed -by `GET https://api.stdcmpt.com/v1/models`. See its -[docs](https://docs.standardcompute.com) for setup. +On [Standard Compute](https://standardcompute.com), pass a model ID exactly +as listed by `GET https://api.stdcmpt.com/v1/models`. The optional +`StandardCompute` router model picks one of your enabled models per request. +The `StandardCompute` router model may route the baseline and candidate to +different underlying models, so results are not reproducible; pin a model ID +for training and evaluation. See its [docs](https://docs.standardcompute.com) +for setup. ### Python API