GPT-6 Sol API Guide: Endpoints, 272K Price Tier, First Call

GPT-6 Sol is in the Router One catalog as openai/gpt-6-sol (observed 2026-09-23): a GPT-6 model with a 1,050,000-token context window, text and image input, and tool calling, served on POST /v1/chat/completions and natively on POST /v1/responses — the wire format Codex CLI speaks. Point any OpenAI-compatible SDK or client at https://api.router.one/v1, send "model": "openai/gpt-6-sol" with a Router One key, and the request is routed, metered and logged like every other model; the model page carries the live rates.

This guide covers what the catalog publishes for the id, how its two price lines work, whether subscription plans cover it, the first request on each endpoint, and what changes when you move traffic over from GPT-6 Astra or GPT-5.6 Sol. Catalog observations are dated; the live catalog is the source of truth for a new request.

Update (2026-10-06): In the 2026-10-06 plan response, GPT-6 Sol and GPT-6.1 Sol are in the Premium models tier of Router One's Pro, Max and Ultra plans; the plan answers in this guide follow that response, and the 2026-10-01 note below records that day's plan response.

Update (2026-10-01): OpenAI released GPT-6.1 Sol on 2026-09-29, and its GPT-6 Sol model page now points to it as the newer Sol model. Router One has listed it as openai/gpt-6.1-sol since 2026-09-30, in no plan tier as of the 2026-10-01 plan response; the GPT-6.1 Sol API guide covers what changes, and GPT-6.1 Sol vs GPT-6 Sol puts the two side by side. GPT-6 Sol stays listed, and OpenAI's deprecations page (checked 2026-10-01) names no retirement for it.

GPT-6 Sol at a glance

OpenAI's own GPT-6 Sol model page adds details the catalog does not carry. It describes the model as "built to power complex coding and agentic workflows", gives an April 20, 2026 knowledge cutoff and a maximum of 922,000 input tokens inside the 1,050,000-token window, and documents reasoning.effort values none, low, medium (the default), high, xhigh and max. Those are the vendor's statements; what Router One vouches for is the catalog entry and how the gateway routes and bills it.

Where it sits in the GPT lineup on Router One

On 2026-09-23 the catalog listed 62 ids. The default-channel GPT chat ids were GPT-6 Astra, GPT-6 Sol, GPT-5.6 Sol, GPT-5.6 Terra and GPT-5.5 (as of the 2026-10-15 catalog, Router One no longer lists GPT-5.5); codex-auto-review, an id Codex's Auto-review may request (per the openai/codex source, rust-v0.163.0, checked 2026-10-17), was listed on its own. The five share the 1,050,000-token window and the GPT endpoint family; all but GPT-5.5, which the catalog then listed with text-only input, also take image input and carry the same 272,000-token tier threshold. What differs is the rate on each model page and how each model does on your prompts — this post does not rank them. GPT-6.1 Sol joined on 2026-09-30 and has its own guide (see the update above).

Three comparison pages render the spec sheets and live rates side by side:

• GPT-6 Sol vs GPT-6 Astra — the two GPT-6 ids listed on 2026-09-23.
• GPT-6 Sol vs GPT-5.6 Sol — generation over generation.
• GPT-6 Sol vs Claude Sonnet 5 — the cross-vendor question.

Two things were not listed on 2026-09-23: GPT-6 Luna, and a channel twin of GPT-6 Sol. The Azure channel carried azure/gpt-6-astra but no azure/gpt-6-sol; the channel model ids guide explains what a channel prefix changes. As of the 2026-10-17 catalog, Router One lists azure/gpt-6-sol, the Azure channel id of the same base model and a separately priced product with its own rates; no plan tier lists azure/gpt-6-sol in the 2026-10-17 plan response, and GPT-6 Luna is still not listed. As of the 2026-10-30 catalog, Router One lists GPT-6 Luna as openai/gpt-6-luna, and the Azure channel lists azure/gpt-6-luna, a separately priced product with its own rates. Ids come and go, so check the catalog before hard-coding one.

How the two price lines work

The model page shows two rate lines, and the total input tokens of one request select which one applies. At exactly 272,000 input tokens the standard line still applies. Strictly above it, the long-context line applies to the whole request — output included — not only to the tokens past the threshold. Monthly volume plays no part. OpenAI's model page documents the same 272K threshold and the same full-request rule for its own list prices, and the Router One tier mirrors it.

That line matters more than the headline window suggests. An agent loop resends its instructions, the conversation and every earlier tool result on each step, so input grows with every turn; a long coding session can cross 272,000 input tokens well before it gets anywhere near 1,050,000. To stay under the line, compact or trim history before it gets there, and read usage on each response to see where a session stands.

The model page also lists cached-input lines. The prompt caching guide shows how to read cache counts in usage so input is not counted twice, the pricing methodology spells out the tier rule, and the cost calculator applies it per request. Reasoning output is billed at the model's posted output rate, with no separate reasoning line (pricing facts).

Do subscription plans cover GPT-6 Sol?

Yes, as of the 2026-10-06 plan response: GPT-6 Sol is in the Premium models tier of Router One's Pro, Max and Ultra plans, beside GPT-6.1 Sol and GPT-5.6 Sol, while GPT-6 Astra sits in the Flagship models tier that only Max and Ultra carry. The tier's allowance is one pool shared by every model it lists; a GPT-6 Sol request draws 1 request, or 2 when its total input is strictly above 272,000 tokens (cached input counts once, output does not count), and calls beyond the cycle's allowance bill the wallet at the posted rates. Moving a Codex session from GPT-5.6 Sol to GPT-6 Sol therefore stays on the same allowance. Plan model lists change, and the pricing page shows the current ones.

Send the first request

1. Create a key. Dashboard → API Keys → Create Key. Keys look like sk-.... For a trial, give the key a maxSpend cap so a runaway loop's wallet charges stop at a number you chose (per-key cost tracking).
2. Set the base URL to https://api.router.one/v1 in any OpenAI-compatible SDK or client.
3. Call either endpoint.

Chat Completions:

curl https://api.router.one/v1/chat/completions \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-6-sol",
    "messages": [{"role": "user", "content": "List three risks of a whole-request price tier for an agent loop."}]
  }'

Responses, with the reasoning effort set explicitly (medium is the vendor default):

curl https://api.router.one/v1/responses \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-6-sol",
    "reasoning": {"effort": "medium"},
    "input": "List three risks of a whole-request price tier for an agent loop."
  }'

The same call from the OpenAI Python SDK — only base_url and the key change:

from openai import OpenAI

client = OpenAI(
    base_url="https://api.router.one/v1",
    api_key="sk-your-api-key",
)

response = client.responses.create(
    model="openai/gpt-6-sol",
    reasoning={"effort": "medium"},
    input="List three risks of a whole-request price tier for an agent loop.",
)
print(response.output_text)
print(response.usage)

stream: true works on both endpoints (streaming guide). For tool calling, OpenAI's GPT-6 Sol model page (checked 2026-10-01) points to the Responses API: Chat Completions supports function calling only with reasoning_effort set to none (tool calling guide). Sending the id to POST /v1/messages returns HTTP 400 invalid_request_error before any model is called, and the message names the paths that serve it. The OpenAI-compatible API page covers the request shape in general, and the API compatibility facts state the endpoint rule per family.

4. Read the trace. Dashboard → Logs shows each call with model, input and output tokens, cost, latency and status. Retryable upstream failures are absorbed by automatic fallback across the candidate routes for the same model (per-request observability).

Codex CLI in one line

If Codex CLI already talks to Router One, the switch is model = "gpt-6-sol" in ~/.codex/config.toml — the bare name, which the gateway resolves to openai/gpt-6-sol — with wire_api = "responses" on the provider as before. This setup is for plan holders (Pro, Max, Ultra), where GPT-6 Sol counts against the Premium models allowance (as of the 2026-10-06 plan response). For a fresh setup, the Codex CLI in China page has the full config, the Codex and Responses API page explains why relays without the Responses wire format return 404, and GPT-6 Sol in Codex CLI, Cursor, Cline and OpenCode walks through the coding clients one by one.

Moving traffic from GPT-6 Astra or GPT-5.6 Sol

For most requests, the move is a model string change: the same key, base URL, endpoints, 1,050,000-token window and 272,000-token threshold. What does change:

• Rates. Each id has its own lines on its model page; the comparison pages above put them side by side on the live catalog.
• Plan coverage. In the 2026-10-06 plan response, GPT-6 Sol and GPT-5.6 Sol share the Premium models allowance of all three plans, and GPT-6 Astra is in the Flagship models tier of Max and Ultra.
• Request rules. Per OpenAI's GPT-6 Sol model page (checked 2026-10-01), Chat Completions supports function calling for GPT-6 Sol only with reasoning_effort set to none, so GPT-5.6 Sol code that sends tools on Chat Completions at another effort should move to the Responses API.
• Behavior on your prompts. A new model answers differently. Run your own prompt set before switching a production path, not a benchmark someone else chose.

A low-risk way to do it: create a second key with a maxSpend cap, send a slice of real traffic to openai/gpt-6-sol through it, and filter Dashboard → Logs by model to compare cost, latency and status per task against the id you use today. Fallback keeps retrying within the same model's candidate routes; it never swaps GPT-6 Sol for a different model, so the comparison stays clean.

From mainland China

Requests reach api.router.one from mainland China without a VPN, on the same key and base URL. Top up with a card or Alipay through one hosted checkout, or with USDT/USDC on six chains (Tron, BSC, Ethereum, Polygon, Base, Arbitrum). No US credit card required. Pricing starts as low as 10% of official provider list prices (up to 90% off) on select models; the GPT API page and each model page show the rate for a given id.

FAQ

What is the model id for GPT-6 Sol on Router One? openai/gpt-6-sol. Since 2026-09-23 the bare name gpt-6-sol resolves to the same model, which is convenient in clients such as Codex CLI that take a bare model name. The model page at /models/gpt-6-sol carries the live rates and the endpoint list.

Can I use GPT-6 Sol in Claude Code? No. Claude Code sends Anthropic Messages requests to /v1/messages, and that endpoint serves Claude-family and DeepSeek ids only; a GPT id there returns HTTP 400 invalid_request_error before any model is called. Use GPT-6 Sol from Codex CLI, which speaks the Responses wire format — on a Pro, Max or Ultra plan, where GPT-6 Sol counts against the Premium models allowance (as of the 2026-10-06 plan response) — or from a client that sends Chat Completions to https://api.router.one/v1 (per OpenAI's model page, checked 2026-10-01, function calling there is supported only with reasoning_effort set to none).

Does the 272K tier apply only to the tokens above the threshold? No. The total input tokens of one request select the line. Exactly at 272,000 the standard line applies; strictly above it, the long-context line applies to the whole request, output included. Monthly usage does not select the tier.

Is GPT-6 Sol included in Router One subscription plans? Yes, in the Premium models tier: in the 2026-10-06 plan response Pro, Max and Ultra list GPT-6 Sol there, beside GPT-6.1 Sol and GPT-5.6 Sol, so its calls draw from that shared allowance and calls beyond it bill the wallet per token. GPT-6 Astra is in the Flagship models tier of Max and Ultra. The pricing page shows the current plan model lists.

Is GPT-6 Luna available through Router One? Not as of the 2026-09-23 catalog. As of the 2026-10-30 catalog, Router One lists it as openai/gpt-6-luna, and the Azure channel lists azure/gpt-6-luna, a separately priced product with its own rates; each model page carries the live rates. The live catalog at /models is the only source of truth; this guide makes no promise about future listings. The GPT-6 Luna API guide covers its model id, OpenAI's request rules and a first request.

Is there an Azure channel version of GPT-6 Sol? Yes, as of the 2026-10-17 catalog: azure/gpt-6-sol, a separately priced product of the same base model with its own rates and isolated routing, served on /v1/chat/completions and natively on /v1/responses. No plan tier lists azure/gpt-6-sol in the 2026-10-17 plan response (openai/gpt-6-sol stays in the Premium models tier), so its calls bill the wallet at its posted rates. On 2026-09-23 the only GPT-6 id on the Azure channel was azure/gpt-6-astra; the channel model ids guide has the rules.

How much does GPT-6 Sol cost through Router One? This guide prints no per-token rates because they go stale. The model page shows both lines live, and the comparison pages render the gap against GPT-6 Astra, GPT-5.6 Sol and Claude Sonnet 5. Pricing starts as low as 10% of official provider list prices (up to 90% off) on select models.

Next steps

• Open the GPT-6 Sol model page for the live rates, both price lines and the endpoint list.
• Compare it with GPT-6 Astra, GPT-5.6 Sol or Claude Sonnet 5.
• Wire it into your coding tools with GPT-6 Sol in Codex CLI, Cursor, Cline and OpenCode.
• Moving to the newer model? Read the GPT-6.1 Sol API guide.
• Looking at GPT-6 Astra? Read the GPT-6 Astra API guide.
• Catch up on the rest of September's catalog changes in the September 2026 new-model guide.