From f10922fd3d4fa0b0989724da3a19ee7ec8ba2469 Mon Sep 17 00:00:00 2001 From: Nikolas Sturm Date: Sat, 12 Sep 2026 21:08:40 +0200 Subject: [PATCH 1/2] feat(providers): add Inco Inco (inco.ai) is a prepaid OpenAI- and Anthropic-compatible inference provider. Adds 7 model routes, all inheriting lab metadata via base_model with provider-specific cost and reasoning controls: - deepseek-v4.1-flash:fast ($0.60/$2.40) - glm-5.3 ($1.40/$4.40), glm-5.3:fast ($2.80/$8.80) - glm-5.3-flash:fast ($0.15/$0.50) - kimi-k3:fast ($6.00/$30.00) - minimax-m3 ($0.30/$1.20), minimax-m3:fast ($0.60/$2.40) Reasoning controls verified against the live API: - GLM models toggle via chat_template_kwargs.enable_thinking - DeepSeek and Kimi toggle via chat_template_kwargs.thinking - reasoning_effort and thinking.type are accepted but ignored - MiniMax M3 always reasons; no caller control (reasoning_options = []) All models return interleaved reasoning in reasoning_content. Logo is the "inco" Montserrat Alternates SemiBold wordmark knocked out of a square, in currentColor. --- providers/inco/logo.svg | 3 +++ .../inco/models/deepseek-v4.1-flash:fast.toml | 15 +++++++++++++++ providers/inco/models/glm-5.3-flash:fast.toml | 15 +++++++++++++++ providers/inco/models/glm-5.3.toml | 15 +++++++++++++++ providers/inco/models/glm-5.3:fast.toml | 13 +++++++++++++ providers/inco/models/kimi-k3:fast.toml | 14 ++++++++++++++ providers/inco/models/minimax-m3.toml | 13 +++++++++++++ providers/inco/models/minimax-m3:fast.toml | 12 ++++++++++++ providers/inco/provider.toml | 5 +++++ 9 files changed, 105 insertions(+) create mode 100644 providers/inco/logo.svg create mode 100644 providers/inco/models/deepseek-v4.1-flash:fast.toml create mode 100644 providers/inco/models/glm-5.3-flash:fast.toml create mode 100644 providers/inco/models/glm-5.3.toml create mode 100644 providers/inco/models/glm-5.3:fast.toml create mode 100644 providers/inco/models/kimi-k3:fast.toml create mode 100644 providers/inco/models/minimax-m3.toml create mode 100644 providers/inco/models/minimax-m3:fast.toml create mode 100644 providers/inco/provider.toml diff --git a/providers/inco/logo.svg b/providers/inco/logo.svg new file mode 100644 index 00000000000..e60b4418f08 --- /dev/null +++ b/providers/inco/logo.svg @@ -0,0 +1,3 @@ + + + diff --git a/providers/inco/models/deepseek-v4.1-flash:fast.toml b/providers/inco/models/deepseek-v4.1-flash:fast.toml new file mode 100644 index 00000000000..2b762eedb7e --- /dev/null +++ b/providers/inco/models/deepseek-v4.1-flash:fast.toml @@ -0,0 +1,15 @@ +# Toggle: chat_template_kwargs.thinking = true|false (verified against Inco; +# reasoning_effort and thinking.type are accepted but ignored). Reasoning is +# off by default on this route; send thinking=true to enable it. +# reasoning_content interleaved field verified against Inco responses. +base_model = "deepseek/deepseek-v4.1-flash" + +[[reasoning_options]] +type = "toggle" + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0.6 +output = 2.4 diff --git a/providers/inco/models/glm-5.3-flash:fast.toml b/providers/inco/models/glm-5.3-flash:fast.toml new file mode 100644 index 00000000000..e4b52c4169f --- /dev/null +++ b/providers/inco/models/glm-5.3-flash:fast.toml @@ -0,0 +1,15 @@ +# Toggle: chat_template_kwargs.enable_thinking = true|false (verified against +# Inco; reasoning_effort and thinking.type are accepted but ignored). Reasons +# by default; reasoning_content interleaved field verified against Inco +# responses. +base_model = "zhipuai/glm-5.3-flash" + +[[reasoning_options]] +type = "toggle" + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0.15 +output = 0.5 diff --git a/providers/inco/models/glm-5.3.toml b/providers/inco/models/glm-5.3.toml new file mode 100644 index 00000000000..bfc8998ba01 --- /dev/null +++ b/providers/inco/models/glm-5.3.toml @@ -0,0 +1,15 @@ +# Toggle: chat_template_kwargs.enable_thinking = true|false (verified against +# Inco; reasoning_effort and thinking.type are accepted but ignored). Reasons +# by default; reasoning_content interleaved field verified against Inco +# responses. +base_model = "zhipuai/glm-5.3" + +[[reasoning_options]] +type = "toggle" + +[interleaved] +field = "reasoning_content" + +[cost] +input = 1.4 +output = 4.4 diff --git a/providers/inco/models/glm-5.3:fast.toml b/providers/inco/models/glm-5.3:fast.toml new file mode 100644 index 00000000000..f92a3d2ec4b --- /dev/null +++ b/providers/inco/models/glm-5.3:fast.toml @@ -0,0 +1,13 @@ +# Standard-speed endpoint at 2x credit pricing, see glm-5.3.toml for controls. +base_model = "zhipuai/glm-5.3" +name = "GLM-5.3 Fast" + +[[reasoning_options]] +type = "toggle" + +[interleaved] +field = "reasoning_content" + +[cost] +input = 2.8 +output = 8.8 diff --git a/providers/inco/models/kimi-k3:fast.toml b/providers/inco/models/kimi-k3:fast.toml new file mode 100644 index 00000000000..ebe03653543 --- /dev/null +++ b/providers/inco/models/kimi-k3:fast.toml @@ -0,0 +1,14 @@ +# Toggle: chat_template_kwargs.thinking = true|false (verified against Inco; +# reasoning_effort and thinking.type are accepted but ignored). Reasons by +# default; reasoning_content interleaved field verified against Inco responses. +base_model = "moonshotai/kimi-k3" + +[[reasoning_options]] +type = "toggle" + +[interleaved] +field = "reasoning_content" + +[cost] +input = 6 +output = 30 diff --git a/providers/inco/models/minimax-m3.toml b/providers/inco/models/minimax-m3.toml new file mode 100644 index 00000000000..dee42c58f8c --- /dev/null +++ b/providers/inco/models/minimax-m3.toml @@ -0,0 +1,13 @@ +# Always reasons on Inco; no caller control — toggle fields (thinking.type, +# chat_template_kwargs.thinking/enable_thinking, reasoning) were tested and are +# ignored. reasoning_content interleaved field verified against Inco responses. +base_model = "minimax/MiniMax-M3" + +reasoning_options = [] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0.3 +output = 1.2 diff --git a/providers/inco/models/minimax-m3:fast.toml b/providers/inco/models/minimax-m3:fast.toml new file mode 100644 index 00000000000..02e06cf191d --- /dev/null +++ b/providers/inco/models/minimax-m3:fast.toml @@ -0,0 +1,12 @@ +# Standard-speed endpoint at 2x credit pricing, see minimax-m3.toml for controls. +base_model = "minimax/MiniMax-M3" +name = "MiniMax M3 Fast" + +reasoning_options = [] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0.6 +output = 2.4 diff --git a/providers/inco/provider.toml b/providers/inco/provider.toml new file mode 100644 index 00000000000..1cbf0c9ba78 --- /dev/null +++ b/providers/inco/provider.toml @@ -0,0 +1,5 @@ +name = "Inco" +npm = "@ai-sdk/openai-compatible" +env = ["INCO_API_KEY"] +api = "https://api.inco.ai/v1" +doc = "https://platform.inco.ai/docs" From c7321d529e1cf6a813e2fd435ae159d634364855 Mon Sep 17 00:00:00 2001 From: Nikolas Sturm Date: Sat, 12 Sep 2026 21:27:25 +0200 Subject: [PATCH 2/2] fix(providers): document Inco toggle wire path on glm-5.3:fast --- providers/inco/models/glm-5.3:fast.toml | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/providers/inco/models/glm-5.3:fast.toml b/providers/inco/models/glm-5.3:fast.toml index f92a3d2ec4b..63d16c7dfe8 100644 --- a/providers/inco/models/glm-5.3:fast.toml +++ b/providers/inco/models/glm-5.3:fast.toml @@ -1,4 +1,6 @@ -# Standard-speed endpoint at 2x credit pricing, see glm-5.3.toml for controls. +# Standard-speed endpoint at 2x credit pricing. +# Toggle: chat_template_kwargs.enable_thinking = true|false +# Verified against Inco; reasons by default. base_model = "zhipuai/glm-5.3" name = "GLM-5.3 Fast"