Kimi K3: Moonshot's 2.8T-Parameter Open Model, Now in the API
Moonshot AI released Kimi K3 this week and it is available through the Text-Generator.io chat completions API as kimi-k3. It is the largest open-weights model released to date: 2.8 trillion total parameters with a 1,048,576-token context window and no length tiering.
Why Kimi K3?
- 1M token context at one flat price -- entire codebases and document sets in a single request
- Vision input and strong tool calling for agent workloads
- Deep reasoning (currently fixed at max effort; more levels coming per Moonshot)
- Open weights -- self-host later with no prompt rework
- #1 debut on the Frontend Code Arena, ahead of Claude and GPT
Pricing through the API is $3.00 per million input tokens and $15.00 per million output tokens, matching Moonshot's official rates.
Using Kimi K3
The endpoint is OpenAI-compatible, so existing clients only change the model name.
Python
from openai import OpenAI
client = OpenAI(
base_url="https://api.text-generator.io/v1",
api_key=API_KEY,
)
response = client.chat.completions.create(
model="kimi-k3",
messages=[{"role": "user", "content": "Review this 400-file repo and list the risky modules"}],
max_tokens=30000,
)
print(response.choices[0].message.content)curl
curl https://api.text-generator.io/v1/chat/completions \
-H "Authorization: Bearer $API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "kimi-k3", "messages": [{"role": "user", "content": "hello"}]}'The alias kimi also resolves to K3. Requests route to Moonshot's API directly with automatic failover to OpenRouter and Kimi K2.5, so one upstream outage never breaks your integration.
One-shot SVG test
Our standing benchmark for new models is a one-shot SVG drawing. K3 was asked for a pelican warrior riding a bicycle -- bronze helmet, spear, flat vector style -- in a single completion:
Practical note: K3 truncates rather than compresses under tight output caps. It overran an 8K completion cap mid-path, then finished this scene cleanly in ~7K tokens when given a 30K budget. Set max_tokens generously for structured artifacts.
Where it fits
Use kimi-k3 for long-context code review, multi-document analysis, and long-horizon tool-calling agents. For high-volume cheap generation keep using the fast tier models -- and let the auto router (text-generator/auto) pick per request.