Recent Announcements
Taiwan payment & e-invoice now live
You can now pay directly on the Buy Credits page; a Taiwan e-invoice is issued automatically after payment.
2026-03-10
New model: DeepSeek V4.1 Flash — 1M context with image input, $0.3 in / $1.2 out
DeepSeek V4.1 Flash is now live on BazaarLink. It is a sparse mixture-of-experts model and the first in the series built on DeepSeek's Causal Encoder-Decoder (CED) architecture, activating 8B parameters on input and 16B on output.
**Specs**: 1,048,576-token context, up to 384,000 output tokens per request. Accepts **text and image input** (text output), and supports tools, structured outputs, response_format and seed.
**Pricing** (per 1M tokens): $0.30 input / $1.20 output / $0.006 cache read. Same context as deepseek-v4-flash in the same series, with image input added.
**Model id**: `deepseek-v4.1-flash` (the fully qualified `deepseek/deepseek-v4.1-flash` also works).
**Note**: this is a reasoning model. It emits reasoning tokens by default and they are billed as output. Leave enough max_tokens for a short answer (too small a budget can be consumed entirely by reasoning, returning an empty string), or use reasoning_effort to bound it.
2026-09-10
New models: GPT Image 2.5 (flare / sunburst) — OpenAI next-gen image generation, $30/M image output
OpenAI's next-generation image models are live on BazaarLink as two variants, `gpt-image-2.5-flare` and `gpt-image-2.5-sunburst`, with the same price and interface: text + image input, image output, 400,000-token context. Pricing per million tokens: $8 input, $8 text output, $30 image output, $2 cache read. Call them via /v1/images/generations or /v1/chat/completions with modalities:["image"]; 1K and 2K sizes are supported today. See the blog post "GPT Image 2.5 API pricing in TWD (2026)" for per-image cost estimation, curl examples and TWD rates.
2026-09-09
Agent self-registration and self-issued API keys are discontinued
As of 2026-09-05, BazaarLink has discontinued the AI-agent self-registration endpoint (/api/v1/agents/register now returns 410), and every agent key that endpoint issued has been deactivated. Requests made with those keys to /api/v1/models or any other endpoint now receive 401.
This closes the source of large-scale automated sign-ups against the free tier and does not affect regular accounts. If you are wiring an agent or automation into BazaarLink, sign up for a regular account and create an API key from the dashboard; existing agent claim links (/claim) still work for claiming an unclaimed account. Questions: reach us via in-app support.
2026-09-06
New model: GPT-6 Astra — 1M-context reasoning model, $10 in / $50 out
OpenAI's next-generation reasoning model GPT-6 Astra is now live on BazaarLink, positioned for computer use, software engineering and long-running agent tasks, with a 1,050,000-token context window and 128,000 max output tokens. Pricing is $10 input / $50 output per 1M tokens (cache read $1, cache write $12.5); requests with more than 272K input tokens are billed at $20 input / $75 output for the whole request. Use model id `gpt-6-astra`; reasoning is always on, temperature is not supported, so control cost with reasoning_effort and leave enough max_tokens for reasoning. See the blog post "GPT-6 Astra API 台幣費用試算 2026 — 1M 上下文、長上下文加價與 GPT-5.5 比較" for the full TWD price table, GPT-5.5 comparison and a curl example.
2026-09-05
