# Telnyx Inference – Kimi K3 Frontier Thinking Model (upto/metered)

> Telnyx Inference – Kimi K3 Frontier Thinking Model (upto/metered) is a paid API for AI agents from x402.telnyx.com, paid per call via x402, $0.408886/call, status unknown (last checked 2026-10-15).

Runs frontier-tier thinking-model chat completions using Moonshot AI's Kimi K3 via Telnyx Inference, billed per-token up to a ceiling, with hidden reasoning tokens counted toward max_tokens.

## Facts

- Endpoint: POST https://x402.telnyx.com/v1/chat/completions/_t/kimi-k3/upto?utm_source=zero.xyz
- Price: $0.408886/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-15
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/telnyx-inference-kimi-k3-frontier-thinking-model-upto-metered-b6911e89
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_VzTlqlOOAaaMMtEHh6rqz

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability telnyx-inference-kimi-k3-frontier-thinking-model-upto-metered-b6911e89 -d '<json body>'
```

Example prompt: Use Kimi K3 on Telnyx to reason through this multi-step problem: given a list of business constraints, propose the optimal staffing schedule — use up to 8192 tokens so the model can think it through fully.

## When to prefer this

Choose this endpoint when you need frontier-tier multi-step reasoning from Kimi K3 and want to pay only for tokens used (metered/upto) rather than a fixed per-call price. Ideal for complex problems — math, code, logic, planning — that benefit from extended hidden reasoning. Prefer this over fixed-price routes when context length and token usage vary significantly across calls. Prefer over cheaper models (DeepSeek Flash, Gemma) when task complexity demands state-of-the-art reasoning quality.

## Known failure modes

- max_tokens below 4096 or above 16384 causes validation error — reasoning tokens require at least 4096
- Context window exceeded if messages array is too large (each message capped at 200,000 chars, max 32 messages)
- Payment failure if USDC balance insufficient for metered token cost
- Model overload or timeout on very long reasoning chains
- Tool array exceeding 16 items returns a validation error
- Temperature out of range (must be 0–2) causes rejection

## How this service works

Frontier thinking-model chat completions via Telnyx Inference (Kimi K3) — hidden reasoning tokens count toward max_tokens, so max_tokens must be 4096–8192 (omit it for the full 8192-token budget). OpenAI-compatible. $0.150193 USDC per call, or pay per-token via the upto scheme. (upto metered — large context)

## Output

Returns an OpenAI-compatible chat completion JSON object containing the assistant's reply (and any tool calls), with hidden reasoning tokens counted in the usage totals. The response includes the generated text, finish reason, token usage breakdown, and tool call payloads if tools were provided.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "n": {
   "type": "integer",
   "const": 1,
   "description": "Fixed to one completion per paid call."
  },
  "model": {
   "type": "string",
   "const": "moonshotai/Kimi-K3",
   "description": "Model identifier; fixed to moonshotai/Kimi-K3 on this route."
  },
  "tools": {
   "type": "array",
   "items": {
    "type": "object"
   },
   "maxItems": 16
  },
  "messages": {
   "type": "array",
   "items": {
    "type": "object",
    "required": [
     "role",
     "content"
    ],
    "properties": {
     "role": {
      "enum": [
       "system",
       "user",
       "assistant",
       "tool"
      ],
      "type": "string"
     },
     "content": {
      "type": "string",
      "maxLength": 200000
     }
    }
   },
   "maxItems": 32,
   "description": "OpenAI-style chat messages."
  },
  "max_tokens": {
   "type": "integer",
   "maximum": 16384,
   "minimum": 4096
  },
  "temperature": {
   "type": "number",
   "maximum": 2,
   "minimum": 0
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/telnyx-inference-kimi-k3-frontier-thinking-model-upto-metered-b6911e89/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from x402.telnyx.com](https://www.zero.xyz/host/x402.telnyx.com/llms.txt)