From ff56555d124c1cfb429fe2aada5c5918f5384def Mon Sep 17 00:00:00 2001 From: JI4JUN Date: Mon, 14 Sep 2026 10:18:12 +0800 Subject: [PATCH] feat(302ai): add deepseek-flash (DeepSeek-V4.1-Flash) 302.AI serves DeepSeek-V4.1-Flash as `deepseek-flash`. Cost from the live catalog ($0.15/$0.60 per MTok, $0.003 cache-read). Reasoning controls verified live: thinking.type toggle suppresses reasoning_content; effort is validated (low|high|max, matching DeepSeek first-party mapping); reasoning_content side channel confirmed. --- providers/302ai/models/deepseek-flash.toml | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) create mode 100644 providers/302ai/models/deepseek-flash.toml diff --git a/providers/302ai/models/deepseek-flash.toml b/providers/302ai/models/deepseek-flash.toml new file mode 100644 index 00000000000..0e607680e15 --- /dev/null +++ b/providers/302ai/models/deepseek-flash.toml @@ -0,0 +1,19 @@ +# DeepSeek-V4.1-Flash on 302.AI, served as `deepseek-flash` (catalog SKU created 2026-09-08). +# Cost: 302.AI catalog, $0.150 / $0.600 per MTok with $0.003 cache-read, https://api.302.ai/v1/models (retrieved 2026-09-14). +# Toggle and `reasoning_content` side channel verified live against POST https://api.302.ai/v1/chat/completions (2026-09-14): +# `thinking.type = disabled` suppresses reasoning output; invalid `reasoning_effort` values are rejected (-10003). +# Toggle: thinking.type = enabled|disabled +# Effort: reasoning_effort = low|high|max +base_model = "deepseek/deepseek-v4.1-flash" +reasoning_options = [ + { type = "toggle" }, + { type = "effort", values = ["low", "high", "max"] }, +] + +[interleaved] +field = "reasoning_content" + +[cost] +input = 0.15 +output = 0.6 +cache_read = 0.003