10B active parameters doing near-flagship work — the cheapest serious model on the market, served at half MiniMax's own price. Built for bulk pipelines, classification, drafts, and agents that run all day.
Exact versions named; vendor-reported figures (MiniMax release + Moonshot compilation, Apr–Jun 2026).
Single-file bug fixes on real repositories — within 3 points of the flagship at ~40× lower output price.
The harder, contamination-resistant variant — M2.7 lands within 1.5 points of GPT-5.4 and ahead of Claude Opus 4.6.
| Provider | 50M in + 10M out / month |
|---|---|
| GPT-5.5 ($5 / $30) | $550.00 |
| Claude Opus 4.8 ($5 / $25) | $500.00 |
| MiniMax M2.7 at its vendor ($0.30 / $1.20) | $27.00 |
| MiniMax M2.7 at Khidi ($0.15 / $0.60) | $13.50 |
client = OpenAI( base_url="https://api.khidi.ai/v1", api_key="sk-…", ) client.chat.completions.create( model="minimax-m2.7", messages=[{"role": "user", "content": "gamarjoba"}], )
OpenAI-compatible with streaming and function calling. Prepaid budgets — your key stops at the limit you set, never a surprise bill.
Khidi also speaks the Anthropic Messages format — drive MiniMax M2.7 from Claude Code:
export ANTHROPIC_BASE_URL="https://api.khidi.ai" export ANTHROPIC_AUTH_TOKEN="sk-…" claude --model minimax-m2.7
Sign up, create a key, start calling — no card required.
Get your API key →