feat(cheapaimodel): add cheapaimodel provider

providers/cheapaimodel/models/deepseek-v4-1-flash.toml

@@ -0,0 +1,30 @@
+# Price: https://cheapaimodel.ai/api/models (fetched 2026-11-02) lists the peak rate
+# (input $0.2100153, output $0.8400612, cache read $0.004200306 per 1M tokens).
+# Billing follows DeepSeek's peak/off-peak split: weekdays (Asia/Shanghai) 09:00-12:00 and
+# 14:00-18:00 Beijing time (01:00-04:00, 06:00-10:00 UTC) bill 2x the off-peak base below.
+# Unlike DeepSeek's own API, Chinese public holidays are not exempted.
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max (mapped from providers/deepseek)
+# Cache: automatic upstream prefix cache, no write price.
+# Live 2026-10-27: reasoning returned in reasoning_content; thinking.type = disabled gives 0 reasoning tokens.
+schema = 2
+base_model = "deepseek/deepseek-v4.1-flash"
+
+[capabilities]
+reasoning = { supported = true, toggle = true, effort = ["low", "high", "max"] }
+structured_output = true
+cache = { implicit = true }
+
+[api.chat-completions]
+compatibility = { reasoning_field = "reasoning_content" }
+
+[cost.text.input]
+price = 0.10500765
+cache = { read = 0.002100153 }
+
+[cost.text.output]
+price = 0.4200306
+
+[[cost.multipliers]]
+when = { utc_days = { in = ["mon", "tue", "wed", "thu", "fri"] }, any = [{ all = [{ utc_hours = { gte = "01:00" } }, { utc_hours = { lt = "04:00" } }] }, { all = [{ utc_hours = { gte = "06:00" } }, { utc_hours = { lt = "10:00" } }] }] }
+factor = 2

providers/cheapaimodel/models/glm-5.3-flash.toml

@@ -0,0 +1,21 @@
+# Price: https://cheapaimodel.ai/api/models (fetched 2026-11-02).
+# Effort: reasoning_effort = low|high|max (mapped from providers/zai)
+# Cache: automatic upstream prefix cache, no write price.
+# Live 2026-10-27: reasoning returned in reasoning_content.
+schema = 2
+base_model = "zhipuai/glm-5.3-flash"
+
+[capabilities]
+reasoning = { supported = true, effort = ["low", "high", "max"] }
+structured_output = true
+cache = { implicit = true }
+
+[api.chat-completions]
+compatibility = { reasoning_field = "reasoning_content" }
+
+[cost.text.input]
+price = 0.10500765
+cache = { read = 0.02100153 }
+
+[cost.text.output]
+price = 0.3500255

providers/cheapaimodel/models/glm-5.3.toml

@@ -0,0 +1,21 @@
+# Price: https://cheapaimodel.ai/api/models (fetched 2026-11-02).
+# Effort: reasoning_effort = low|high|max (mapped from providers/zai)
+# Cache: automatic upstream prefix cache, no write price.
+# Live 2026-10-27: reasoning returned in reasoning_content.
+schema = 2
+base_model = "zhipuai/glm-5.3"
+
+[capabilities]
+reasoning = { supported = true, effort = ["low", "high", "max"] }
+structured_output = true
+cache = { implicit = true }
+
+[api.chat-completions]
+compatibility = { reasoning_field = "reasoning_content" }
+
+[cost.text.input]
+price = 0.9800714
+cache = { read = 0.18201326 }
+
+[cost.text.output]
+price = 3.0802244

providers/cheapaimodel/models/grok-4.7.toml

@@ -0,0 +1,20 @@
+# Price: https://cheapaimodel.ai/api/models (fetched 2026-11-02). One flat rate at every
+# context length (no long-context tier) and no priority tier on this host.
+# Effort: reasoning_effort = low|medium|high|xhigh (mapped from providers/xai)
+# Live 2026-10-27 on /v1/chat/completions: reasoning text is not returned (reasoning_content is
+# null), only usage.completion_tokens_details.reasoning_tokens, which rises with effort (low 546, high 680).
+# APIs: Chat Completions only on this host.
+schema = 2
+base_model = "xai/grok-4.7"
+
+[capabilities]
+reasoning = { supported = true, effort = ["low", "medium", "high", "xhigh"] }
+structured_output = true
+cache = { implicit = true }
+
+[cost.text.input]
+price = 0.199588
+cache = { read = 0.049897 }
+
+[cost.text.output]
+price = 0.598764

providers/cheapaimodel/models/kimi-k3.toml

@@ -0,0 +1,21 @@
+# Price: https://cheapaimodel.ai/api/models (fetched 2026-11-02).
+# Effort: reasoning_effort = low|high|max (mapped from providers/moonshotai); always thinks.
+# Cache: automatic upstream prefix cache; the catalog lists a cache-write price.
+# Live 2026-10-27: reasoning returned in reasoning_content.
+schema = 2
+base_model = "moonshotai/kimi-k3"
+
+[capabilities]
+reasoning = { supported = true, effort = ["low", "high", "max"] }
+structured_output = true
+cache = { implicit = true }
+
+[api.chat-completions]
+compatibility = { reasoning_field = "reasoning_content" }
+
+[cost.text.input]
+price = 2.100153
+cache = { read = 0.2100153, write = 2.100153 }
+
+[cost.text.output]
+price = 10.500765

providers/cheapaimodel/provider.toml

@@ -0,0 +1,14 @@
+# cheapaimodel is an OpenAI-compatible relay: it resells first-party models
+# (DeepSeek, Z.ai, xAI, Moonshot) through one key, billed in USD per token.
+# Public catalog with prices, limits and modalities (no auth):
+#   https://cheapaimodel.ai/api/models (fetched 2026-10-27)
+# Prices are list price times a per-group discount and move with the monthly
+# USD/CNY fixing used by the upstream, so they can shift slightly month to month.
+# Reasoning controls are passed through to the lab and mapped from each lab's
+# first-party entry (not probed per host).
+name = "cheapaimodel"
+npm = "@ai-sdk/openai-compatible"
+opencodeai = "@opencode/ai/providers/openai-compatible"
+env = ["CHEAPAIMODEL_API_KEY"]
+api = "https://api.cheapaimodel.ai/v1"
+doc = "https://cheapaimodel.ai/docs"