DeepSeek V4.1 Flash

DeepSeek V4.1 Flash carries a 1M-token context window and a 384,000-token output ceiling, among the largest in the catalog, and it accepts images as well as text. It reasons before it answers, and those reasoning tokens bill as output: a one-line translation measured 1,427 reasoning tokens against 19 of actual reply, so budget output generously and expect the thinking, rather than the answer, to set the cost of a short response. Tool calling works as sent, with no special settings required. JSON object mode is supported; strict JSON schema is not, so pass a tool definition when you need a guaranteed shape. Arabic is strong across formal writing, translation and Gulf dialects, which it distinguishes by country. The earlier V4 Flash builds stay available under their own ids for anyone who built against them.

Quick start

Use any OpenAI-compatible client. Just change base_url and your key.

from openai import OpenAI

client = OpenAI(
    api_key="tl-xxxxxxxxxxxxxxxxxxxxxxxx",
    base_url="https://api.thalam.ai/v1",
)

response = client.chat.completions.create(
    model="deepseek/deepseek-v4.1-flash",
    messages=[
        {"role": "user", "content": "Hello from Thalam!"}
    ],
)

print(response.choices[0].message.content)