Add deepseek-v4.1-flash to the international Alibaba Token Plan provider

[oliver-mee]

Problem

deepseek-v4.1-flash landed on the Alibaba Token Plan gateway on 2026-09-15. The China provider already carries it thanks to #7045, which used the Beijing-region plan list as its evidence. But the model listed the same day on the international (Singapore) endpoint too, on both the Team and Personal editions. providers/alibaba-token-plan/models/ has no row, so international Token Plan users reading models.dev cannot select a model their key actually serves.

Evidence (checked 2026-09-15)

• Authenticated GET /models on the Singapore endpoint (token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1): deepseek-v4.1-flash present on Team and Personal keys. Beijing endpoint lists it as well (matches feat(alibaba-token-plan-cn): add DeepSeek V4.1 Flash #7045).
• Live completion against the Singapore plan gateway: 200, reasoning streams on reasoning_content.
• enable_thinking: true|false both accepted (hybrid model).
• reasoning_effort: the gateway's own validation error names minimal, low, medium, high, xhigh, max, ultra as the accepted string values. low/high/max are the ladder that behaves per the series documentation, and they match the values feat(alibaba-token-plan-cn): add DeepSeek V4.1 Flash #7045 merged for China, so a global entry consistent with China should carry the same three.
• The lab file models/deepseek/deepseek-v4.1-flash.toml already exists, so the provider entry is override-only (base_model, reasoning_options, interleaved, zeroed subscription cost).

Proposed fix

Add providers/alibaba-token-plan/models/deepseek-v4.1-flash.toml, mirroring the merged China entry #7045 exactly:

base_model = "deepseek/deepseek-v4.1-flash"

reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "high", "max"] }]

[interleaved]
field = "reasoning_content"

[cost]
input = 0
output = 0
cache_read = 0
cache_write = 0

with the source-access comment block in house style pointing at the OpenAI-compatible chat API reference and the international Token Plan personal/team overview pages.

Costs are zeroed because Token Plan is a Credits subscription with no per-token USD rate on this endpoint, same rationale as every other row in both providers.