GPT-6.1 Sol

Last updated: 30 September 2026. Official prices and limits from OpenAI's GPT-6.1 Sol model page.

GPT-6.1 Sol vs GPT-6 Sol, Astra, and Luna

GPT-6.1 Sol is OpenAI's upgrade to GPT-6 Sol, released on 29 September 2026. It sits between GPT-6 Astra, the most capable model in the family, and GPT-6 Luna, the low-cost tier for focused, high-volume tasks.

Base input and output rates are unchanged from GPT-6 Sol. The difference is cached input, which drops from 10% to 5% of the uncached rate. Against Astra, short-context input and output cost one fifth as much. Prices below are per 1M tokens. The AIReiter columns are what you pay here.

What changed from GPT-6 Sol

• Cached input is half the price. Official cache reads drop from $0.20 to $0.10 per 1M, and from $0.06 to $0.03 on AIReiter.
• Reasoning effort gains max. Supported values are low, medium (default), high, xhigh, and max. The none and minimal settings are not supported, so remove them from existing requests before switching the model ID.
• Same limits. A 1,050,000-token context window, 128K output tokens, text and image input, and the same 272K long-context surcharge.

Which one to pick

Use Luna for routine, high-volume traffic such as classification, extraction, and first-line support. Use GPT-6.1 Sol for hard code, long documents, and agents that must recover from failed steps. Keep Astra for the small share of tasks where Sol measurably falls short, since it costs five times as much per token.

Call GPT-6.1 Sol from your code

The endpoint is OpenAI-compatible, so existing OpenAI clients work by changing the base URL and the API key. The model ID is gpt-6.1-sol.

Tool calling works on both endpoints. On OpenAI's own API, GPT-6.1 Sol supports tool calling only through the Responses API, not Chat Completions. On AIReiter, Chat Completions requests with tools are handled for you, so existing tool loops keep working after you switch the model ID. New integrations can use either /api/v1/chat/completions or /api/v1/responses.

curl (Chat Completions)

curl https://aireiter.com/api/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AIREITER_API_KEY" \
  -d '{
    "model": "gpt-6.1-sol",
    "messages": [{"role": "user", "content": "Explain how a 429 response should be retried."}],
    "reasoning_effort": "medium",
    "stream": true
  }'

Python with tool calling (Responses API)

from openai import OpenAI

client = OpenAI(
    base_url="https://aireiter.com/api/v1",
    api_key="YOUR_AIREITER_API_KEY",
)

response = client.responses.create(
    model="gpt-6.1-sol",
    input="What is the weather in Paris right now?",
    reasoning={"effort": "medium"},
    tools=[{
        "type": "function",
        "name": "get_weather",
        "description": "Get the current weather for a city.",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
        },
    }],
)

for item in response.output:
    if item.type == "function_call":
        print(item.name, item.arguments)

Node (Chat Completions, streaming)

import OpenAI from "openai"

const client = new OpenAI({
  baseURL: "https://aireiter.com/api/v1",
  apiKey: process.env.AIREITER_API_KEY,
})

const stream = await client.chat.completions.create({
  model: "gpt-6.1-sol",
  messages: [{ role: "user", content: "Explain how a 429 response should be retried." }],
  reasoning_effort: "medium",
  stream: true,
})

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "")
}

Use GPT-6.1 Sol in Codex CLI

Codex CLI talks to the Responses API, so tool use works out of the box. Put this in ~/.codex/config.toml, then export your key as AIREITER_API_KEY:

model_provider = "aireiter"
model = "gpt-6.1-sol"
model_reasoning_effort = "high"
disable_response_storage = true

[model_providers.aireiter]
name = "AIReiter"
base_url = "https://aireiter.com/api/v1"
wire_api = "responses"
env_key = "AIREITER_API_KEY"

base_url must stop at /api/v1. Keep the real key in the environment variable, never in the TOML file. Grab a key on the API keys page and see the LLM API integration guide for other clients.

What GPT-6.1 Sol actually costs you

Here is the arithmetic on three realistic workloads at AIReiter's rate of $0.60 input and $3.00 output per 1M tokens, with no cache hits.

• Cached input reads cost $0.03 per 1M, one twentieth of a fresh read. If 40K of that 50K agent step is a cached prefix, the step drops from $42.00 to about $19.20 per 1,000 requests. A stable system prompt and context prefix matter more on GPT-6.1 Sol than on GPT-6 Sol.
• Requests above 272K input tokens are surcharged for the whole request, at 2x on input and cache and 1.5x on output. A 300K-in / 4K-out request costs about $0.378, while the same task trimmed to 270K costs about $0.174.

Output tokens cost five times what input tokens cost. Higher reasoning effort produces more reasoning tokens, which are billed as output, so reserve xhigh and max for problems that need them.