# Effort: reasoning_effort = low|high|max (DeepSeek flash maps low→low; xhigh→high)
# https://api-docs.deepseek.com/guides/thinking_mode/ (accessed 2026-08-02)
# Mid-conversation flags left out: deepseek has no provider file for deepseek/deepseek-v4-flash-0731 to follow (Zen passes requests through).
# Tools follow DeepSeek's API: https://api-docs.deepseek.com/api/create-chat-completion (tool_choice; strict needs DeepSeek's beta base URL, so it is left out); Zen passes requests through.
schema = 2
base_model = "deepseek/deepseek-v4-flash-0731"
name = "DeepSeek V4 Flash Free"
status = "deprecated"

[capabilities]
tools = { supported = true, choice = ["auto", "none", "required", "tool"] }
reasoning = { supported = true, effort = ["low", "high", "max"] }
structured_output = true

[limit]
context = 200_000
output = 128_000

[api.chat-completions]
base_url = "https://opencode.ai/inference/openai/v1"
compatibility = { reasoning_field = "reasoning_content" }

[cost.text.input]
price = 0
cache = { read = 0 }

[cost.text.output]
price = 0