Standard Compute

OpenCode · Facts verified 2026-10-14

How do I use Kimi K3 in OpenCode?

Add moonshotai/kimi-k3 to a standardcompute provider in opencode.json with base URL https://api.stdcmpt.com/v1, set it as the model and point small_model at z-ai/glm-5.3-flash: OpenCode runs on Kimi K3 without a Moonshot account, on a plan from $19 a month that returns HTTP 402 to new requests once its compute budget is used.

opencode.jsonchat completions
{
  "$schema": "https://opencode.ai/config.json",
  "model": "standardcompute/moonshotai/kimi-k3",
  "small_model": "standardcompute/z-ai/glm-5.3-flash",
  "provider": {
    "standardcompute": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Standard Compute",
      "options": {
        "baseURL": "https://api.stdcmpt.com/v1",
        "apiKey": "{env:STANDARDCOMPUTE_API_KEY}"
      },
      "models": {
        "moonshotai/kimi-k3": {
          "name": "Kimi K3",
          "reasoning": true,
          "attachment": true,
          "modalities": { "input": ["text", "image"], "output": ["text"] },
          "limit": { "context": 400000, "output": 32000 }
        },
        "z-ai/glm-5.3-flash": {
          "name": "GLM 5.3 Flash",
          "reasoning": true,
          "attachment": true,
          "modalities": { "input": ["text", "image"], "output": ["text"] },
          "limit": { "context": 400000, "output": 32000 }
        }
      }
    }
  }
}

small_model sends session titles and other light calls to GLM 5.3 Flash instead of K3. Merge into your existing opencode.json.

K3 is the costly part of an OpenCode day

K3 lists at $3 input and $15 output per million tokens. At Moonshot's list price, the Starter plan's $20 compute budget covers about 3 typical days and Standard ($89 a month, $95 of compute) about 15.

small_model moves session titles and summaries to GLM 5.3 Flash at $0.15 input; the build turns stay on K3.

Kimi K3 at list priceCost
Typical day$6.16
Busy day$63
20 typical days$123

Kimi K3 list prices per million tokens: $3.00 input on a cache miss, $0.30 on a hit, $15.00 output. Source. A typical day is 8.5M tokens and a busy one 86.7M, 87.2% of input as cache reads: the median and 90th-percentile coding-agent day in our Agent Burn Index, measured on OpenCode, Cline, Kilo and similar agents, September 13 to October 12, 2026. Priced at the list rates above; your own days will differ.

Kimi K3: model ID, providers and price

What works with K3 in OpenCode

  • Tools, edits and MCP servers work over Chat Completions.
  • Screenshots work: K3 reads images, and the config declares it. The US-served moonshotai/kimi-k3:us is text-only.
  • OpenCode has no built-in reasoning variants for Kimi IDs. Set K3's level (minimal to high) on the Models page, or add a variants block with reasoningEffort to the model.

The catch

OpenCode Go at $10 a month includes Kimi K3 with a $15 monthly usage limit at OpenCode's prices ($60 on Go Plus at $40), which OpenCode estimates at about 110 K3 requests per 5 hours on Go. OpenCode also lists Moonshot as a built-in provider (MOONSHOT_API_KEY, moonshotai/kimi-k3). K3 all day is expensive on any route.

Where Standard Compute fits

K3 for the hard turns and a lower-priced model one /models pick away, on one key with no 5-hour or weekly windows and no Moonshot account or prepaid balance.

Get your API key Plans from $19/mo

FAQ

Is Kimi K2.7 Code available on Standard Compute?

No. K3 is the Kimi model served here; GET https://api.stdcmpt.com/v1/models lists every ID. OpenCode Go and Moonshot's own API still offer K2.7 Code.

Which is better in OpenCode, Kimi K3 or GLM 5.3?

GLM 5.3 costs less per day at list price (about $3.65 against $6.16) but cannot read images. Try both on one real task; the same key serves both.

Related setup guides