AIAI SPACE

frontier, but yours

State-of-the-art open models
on infrastructure you trust.

Ask anything…
Kimi K3🇺🇸

Drop-in for OpenCode — and works with Claude Code too.

Start for $25/mo
GLM-5.2 — #1 for web & UI generation on DesignArena

Frontier capability, full data sovereignty.

Kimi K3, GLM-5.2, and Kimi K2.7 Code — the top open coding models — running on Cloudflare's global network across the US, UK, Germany, Japan and Australia. Frontier capability, served close to you, with your code staying on trusted allied infrastructure.

Routed across Cloudflare's global network · US · UK · DE · JP · AU

United States
United Kingdom
Germany
Japan
Australia
Cloudflare's global network
|
Trusted allied infrastructure, full data sovereignty
|
Anthropic-compatible API
Models

Five open models. One subscription.

Three frontier flagships and two specialized models. Switch with a single env var; Kimi K3 is the default.

Kimi K3

Moonshot AI

Default

Moonshot's new flagship and the most capable model on AI Space: a 2.8T-parameter model with a 1M-token context window, native vision, and always-on reasoning you can dial with reasoning_effort.

  • 1M context
  • 2.8T params
  • Native vision
  • Always-on reasoning
sota/kimi-k3

GLM-5.2

Z.ai

Flagship

Z.ai's flagship agentic coding model. Exceptional at long-horizon tasks, multi-file edits, and tool use. 262K context window.

  • 262K context
  • Agentic coding
  • Tool use
  • Extended thinking
#1 for web & UI on DesignArena
sota/glm-5.2

Kimi K2.7 Code

Moonshot AI

Frontier

Moonshot's frontier coding model. State-of-the-art performance on SWE-bench and competitive programming. Optimized for precision.

  • SWE-bench leading
  • Precision coding
  • Competitive perf
  • Fast inference
sota/kimi-k2.7-code

Also on AI Space

Two specialized models for when you'd trade top-end capability for speed, cost, or cheaper agentic reasoning.

GLM-4.7 Flash

Z.ai

Fast & cheap

Z.ai's flash tier: a compact ~30B MoE that still scores 59.2 on SWE-bench Verified — frontier-grade for its size. Reach for it on high-volume, latency-sensitive, or well-scoped edits.

  • 131K context
  • SWE-bench 59.2
  • Fast inference
sota/glm-4.7-flash

Nemotron-3 120B

NVIDIA

Reasoning

NVIDIA's hybrid Mamba-Transformer MoE: 120B total, 12.7B active. Built for agentic reasoning and multi-agent workflows, at a cheap-tier price.

  • 256K context
  • 120B / 12.7B active
  • Agentic reasoning
sota/nemotron-3-120b
How it works

Up and running in three steps

01

Subscribe

Choose Starter ($25/mo) or Pro ($125/mo). Cancel anytime.

02

Create an API key

Generate a key from your dashboard in seconds.

03

Connect OpenCode (or Claude Code)

Drop your key into opencode.json — or set two env vars for Claude Code. That's it.

Setup

Built for OpenCode. Works with Claude Code too.

Once you subscribe, grab your key from the dashboard and connect in seconds.

opencode.json
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "sota": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "AI Space",
      "options": {
        "baseURL": "https://ai.xyspace.dev/v1",
        "apiKey": "sk-sota-…"  // from your dashboard
      },
      "models": {
        "kimi-k3": { "name": "Kimi K3" },
        "glm-5.2": { "name": "GLM-5.2" },
        "kimi-k2.7-code": { "name": "Kimi K2.7 Code" }
      }
    }
  }
}
FAQ

Common questions