Moonshot AI's open-weight flagship, served OpenAI-compatible at half the vendor's own price — with streaming, tool calling and 256K context.
Exact model versions named; vendor-reported figures (Moonshot AI release, April 2026).
Real GitHub issues — the contamination-resistant coding benchmark.
The hardest knowledge benchmark, solved the way agents actually work. Only the newest flagship stays ahead.
Deep retrieval and synthesis — the core skill of research agents. +13.9 points over GPT-5.4.
| Provider | 50M in + 10M out / month |
|---|---|
| GPT-5.5 ($5 / $30) | $550.00 |
| Claude Opus 4.8 ($5 / $25) | $500.00 |
| Kimi K2.6 at Moonshot ($0.95 / $4.00) | $87.50 |
| Kimi K2.6 at Khidi ($0.50 / $2.00) | $45.00 |
client = OpenAI( base_url="https://api.khidi.ai/v1", api_key="sk-…", ) client.chat.completions.create( model="kimi-k2.6", messages=[{"role": "user", "content": "gamarjoba"}], )
Works with Cursor, LangChain, n8n, LibreChat — anything OpenAI-compatible. Streaming and function calling included. Prepaid budgets: your key stops at the limit you set.
Khidi also speaks the Anthropic Messages format — drive Kimi K2.6 from Claude Code:
export ANTHROPIC_BASE_URL="https://api.khidi.ai" export ANTHROPIC_AUTH_TOKEN="sk-…" claude --model kimi-k2.6
Sign up, create a key, start calling — no card required.
Get your API key →