New Models Added: GPT-6 Sol, GPT-6 Luna, and Grok 4.7

Date: 2026-09-24 / (1405-07-02)

Summary

OpenAI's GPT-6 Sol and GPT-6 Luna, and xAI's Grok 4.7, are now available on AvalAI for coding, reasoning, and professional work. All three support Chat Completions and Messages. Responses support is full for Sol and Luna and partial for Grok 4.7.

---

Details

OpenAI: GPT-6 Sol and GPT-6 Luna

• gpt-6-sol brings GPT-6 capabilities to cost-conscious coding agents, professional analysis, computer-use workflows, and long conversations. OpenAI reports improvements over GPT-5.6 Sol in factuality, software engineering, and verification.
• gpt-6-luna is the lower-cost option for high-volume assistance, document processing, coding subtasks, and tool-driven workflows. It also supports reasoning and image understanding.
• gpt-6-astra remains OpenAI's highest-capability GPT-6 option for the most demanding tasks; this release adds choices rather than announcing an automatic model replacement.

Both new models support text and image input, text output, reasoning, function calling, structured output, PDF input, and prompt caching. AvalAI lists 922,000 maximum input tokens and 128,000 maximum output tokens for each model. Do not carry over a different model's context limits when migrating.

OpenAI reports improved cache reuse, including when reasoning effort or tool availability changes, and additional controls such as explicit cache breakpoints. These upstream improvements do not establish pass-through support for every cache control on every AvalAI route. Keep reusable prefixes stable, measure actual cache hits, and consult the prompt caching guide. Cached-input token rates below are 90% lower than the corresponding uncached-input rates; cache creation is priced separately.

See the OpenAI provider guide, GPT-6 Sol model page, and GPT-6 Luna model page.

xAI: Grok 4.7

Use grok-4.7 for long-running coding agents, careful verification, document and presentation work, and technical analysis. xAI describes a larger base model, longer training on difficult tasks, better long-context handling, and revised safeguards compared with Grok 4.6. These are provider-reported improvements, not an AvalAI performance guarantee.

AvalAI lists 500,000 maximum input tokens and 500,000 maximum output tokens, with vision, reasoning, tool calling, structured output, and prompt caching. These are the catalog's separate limits, not a promise that both maxima can be used together.

The reference includes higher-effort benchmark results, but those labels do not establish an AvalAI reasoning_effort enum or default. Start without effort overrides and verify any additional controls on your selected route. The separately advertised upstream fast variant and invite-only security features are not announced as AvalAI routes or entitlements here.

See the xAI provider guide and Grok 4.7 model page.

Endpoint availability

Model	v1/chat/completions	v1/messages	v1/responses
gpt-6-sol	Full	Full	Full
gpt-6-luna	Full	Full	Full
grok-4.7	Full	Full	Partial

Partial Responses support is not full hosted-tool or stateful-workflow parity. Test tool round-trips, output formats, and conversation continuation before moving Grok workflows to Responses. Endpoint availability alone also does not enable every hosted tool for every model or account.

Image understanding is not image generation

These three models are not image-generation models. Use them to analyze images or prepare text instructions, not as the image model in v1/images/generations or v1/images/edits. For direct generation and editing, keep using dedicated models such as gpt-image-2.5-flare or gpt-image-2.5-sunburst.

Using the hosted image_generation tool in Responses is a separate capability that must be verified for the chosen model, route, and account. Full Responses endpoint support for Sol and Luna does not by itself confirm that tool, and Grok 4.7's partial support does not imply it. See the Image API and image generation guide.

---

Pricing

All prices below are USD per 1 million tokens, not subscription prices. The input-length threshold determines the applicable input, cache, and output rates for the request. The higher tier applies only when input length exceeds the threshold; it is not a marginal surcharge on just the excess tokens.

GPT-6 Sol and GPT-6 Luna

Model	Input length	Input	Cached input	Cache creation input	Output
gpt-6-sol	Up to and including 272K	$2.00	$0.20	$2.50	$10.00
gpt-6-sol	Above 272K	$4.00	$0.40	$5.00	$15.00
gpt-6-luna	Up to and including 272K	$0.10	$0.01	$0.125	$0.50
gpt-6-luna	Above 272K	$0.20	$0.02	$0.25	$0.75

Grok 4.7

Input length	Input	Cached input	Output
Up to and including 200K	$2.00	$0.50	$6.00
Above 200K	$4.00	$1.00	$12.50

Grok 4.7 uses the same token rates as Grok 4.6. Its above-200K tier is long-context pricing, not pricing for an upstream fast variant. Consult pricing for the full catalog, any separately billed tools, and account-tier rate limits.

---

API request and SDK examples

Set AVALAI_API_KEY in your server environment. The examples use the same minimal Chat Completions request. Change only the model ID to gpt-6-luna or grok-4.7 to compare the other models; do not copy unsupported effort or sampling overrides from an older model.

Bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "gpt-6-sol",
    "messages": [
      {"role": "user", "content": "Suggest a concise verification checklist for deploying an API behind a feature flag."}
    ]
  }'

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"],
    base_url="https://api.avalai.ir/v1",
)

response = client.chat.completions.create(
    model="gpt-6-sol",
    messages=[
        {
            "role": "user",
            "content": "Suggest a concise verification checklist for deploying an API behind a feature flag.",
        }
    ],
)

print(response.choices[0].message.content)

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.AVALAI_API_KEY,
  baseURL: "https://api.avalai.ir/v1",
});

const response = await client.chat.completions.create({
  model: "gpt-6-sol",
  messages: [
    { role: "user", content: "Suggest a concise verification checklist for deploying an API behind a feature flag." },
  ],
});

console.log(response.choices[0].message.content);

Illustrative response

This is an illustrative response, not a live API capture. The identifier, timestamp, token counts, and exchange rate are examples. With 100 uncached input tokens and 50 output tokens, no cache creation, no paid tools, and input below 272K, Sol's token cost is $0.0007. The example exchange rate is 100,000 Tomans per USD, not a current quote. Actual usage, optional fields, and billing depend on the request.

{
  "id": "chatcmpl-gpt-6-sol-example",
  "created": 1790208000,
  "model": "gpt-6-sol",
  "object": "chat.completion",
  "system_fingerprint": null,
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Verify the flag defaults off, test both paths, check authentication and error handling, monitor latency and errors during a limited rollout, and confirm rollback restores the previous behavior.",
        "annotations": []
      }
    }
  ],
  "usage": {
    "prompt_tokens": 100,
    "completion_tokens": 50,
    "total_tokens": 150,
    "completion_tokens_details": null,
    "prompt_tokens_details": {
      "cached_tokens": 0,
      "text_tokens": 100,
      "audio_tokens": null,
      "image_tokens": null
    }
  },
  "estimated_cost": {
    "unit": "0.0007000000",
    "irt": 70,
    "exchange_rate": 100000
  }
}

Responses example for Sol and Luna

For a Responses-first integration, use the following request and replace gpt-6-luna with gpt-6-sol when appropriate. This example does not request any hosted tool. Grok 4.7 still requires workflow-specific testing because its Responses support is partial.

Bash

curl https://api.avalai.ir/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "gpt-6-luna",
    "input": "Give a concise rollout checklist for a new API endpoint."
  }'

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"],
    base_url="https://api.avalai.ir/v1",
)

response = client.responses.create(
    model="gpt-6-luna",
    input="Give a concise rollout checklist for a new API endpoint.",
)

print(response.output_text)

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.AVALAI_API_KEY,
  baseURL: "https://api.avalai.ir/v1",
});

const response = await client.responses.create({
  model: "gpt-6-luna",
  input: "Give a concise rollout checklist for a new API endpoint.",
});

console.log(response.output_text);

---

Integration checklist

1. Choose the explicit model ID: gpt-6-sol, gpt-6-luna, or grok-4.7. No automatic rerouting or retirement of earlier models is announced here.
2. Compare representative prompts, tool results, structured outputs, image inputs, and response quality before switching production traffic.
3. Check input limits, leave output headroom for hidden reasoning tokens, and verify controls against the selected endpoint. Request concise conclusions and verification steps, not hidden chain-of-thought.
4. Measure cache usage and total token cost, including the applicable long-context tier and any separately billed tools.
5. For new Sol and Luna workflows, consider Responses. For Grok 4.7, start with Chat Completions unless your exact Responses workflow has been validated.

Related documentation

• OpenAI models
• xAI models
• Chat Completions, Messages, and Responses
• Reasoning models and prompt caching
• Pricing and model comparison

For integration or billing questions, contact AvalAI support.