How to Use GPT-6.1 Sol: API & Codex Setup

GPT-6.1 Sol Hands-On: From Zero to a Working $2/$10 Agent Call

The fastest way to use GPT-6.1 Sol today is the GPT-6.1 Sol API via kie.ai, which exposes the model at its standard $2 per million input and $10 per million output tokens without an OpenAI subscription. If you are already an OpenAI customer, the same model is reachable through OpenAI's own API under the string gpt-6.1-sol, and inside Codex and ChatGPT Work for Plus, Pro, Business, Enterprise, and Edu plans. The model shipped on September 29, 2026 at DevDay 2026 and is generally available now — there is no waitlist.

Key Takeaways

• GPT-6.1 Sol is live and GA as of September 29, 2026; the API model string is gpt-6.1-sol.
• Standard pricing is $2/M input, $10/M output, and $0.10/M cached input (a 95% cache discount).
• The model holds a 1M-token input context window and up to 128K output tokens.
• It scores 52 on the Artificial Analysis Intelligence Index, one point behind GPT-6 Astra, at under a quarter of Astra's cost per task.
• kie.ai offers GPT-6.1 Sol via API; OpenAI offers it via API, Codex, and ChatGPT Work.
• An Ultrafast tier runs up to 8x faster at $12/M input and $60/M output.

Ways to Access GPT-6.1 Sol

GPT-6.1 Sol is a Chat/agentic model, so every route below is an API or a hosted coding surface. The table lists the two first-class paths: the kie.ai API and OpenAI's own channels.

Channel	Barrier to entry	Cost	Best for
kie.ai API	Sign up, create API key, add billing	$2/M in · $10/M out (standard)	Developers who want the model without an OpenAI subscription
OpenAI API	OpenAI account with billing configured	$2/M in · $10/M out · $0.10/M cached	Teams already on OpenAI infrastructure
Codex (OpenAI)	Plus/Pro/Business/Enterprise/Edu plan	Included in plan usage	Agentic coding in the terminal and IDE
ChatGPT Work	Paid ChatGPT plan	Included in plan usage	Document and workflow tasks inside ChatGPT
Ultrafast tier	API; Codex/Work gated to Pro $500	$12/M in · $60/M out	Latency-sensitive agent loops

Regular ChatGPT chat did not carry GPT-6.1 Sol at launch. As of October 11, 2026, OpenAI has begun rolling GPT-6 Sol into the ChatGPT Chat tab for paid plans, with Luna for Free and Go users.

Using GPT-6.1 Sol via the kie.ai API

For developers who want GPT-6.1 Sol on a metered API without an OpenAI subscription, kie.ai lists it as a first-class model. The steps below are the standard flow; the exact request body lives in the model page docs.

Step 1 — Open the model page

Start at the GPT 6.1 Sol model page on kie.ai. The page documents the modality (Chat), the supported parameters, and the request format. Confirm the model string there before writing any code rather than assuming it matches another provider's naming.

Step 2 — Register and get an API key

Create an account and generate an API key from the dashboard. Attach a billing method; GPT-6.1 Sol is a paid model with no free tier, so an unfunded key will return an auth or quota error on the first call.

Step 3 — Try it in the playground

Run a short prompt in the playground to verify the key works and to feel the model's latency before you wire it into an agent loop. GPT-6.1 Sol measured roughly 20-30 tokens per second for many early users, so a throughput check here saves debugging time later.

Step 4 — Call it from your code

Send your first request following the exact schema on the model page. Because GPT-6.1 Sol supports a 1M-token input window and a 95% cache discount, lean on prompt caching for repeated system prompts and large codebases — cached input drops to $0.10 per million tokens. Check the API docs on the model page for the exact request format, headers, and parameter names.

Using GPT-6.1 Sol in the Official OpenAI App

Inside OpenAI's own products, GPT-6.1 Sol appears in Codex and ChatGPT Work for paid plans. In Codex, select the model from the model picker; one user needed to delete and reinstall the Codex Windows app before the model appeared, which is a reasonable first fix if it is missing.

Source: @HarshithLucky3

Codex exposes reasoning-effort levels including high and xhigh. One tester reported that the $20 Codex plan felt "almost unlimited" for GPT-6.1 Sol work, and a Codex Plus run logged 40.8M tokens before hitting a session limit in about two hours. On the API side, GPT-6.1 Sol also surfaced in GitHub Copilot and Microsoft Foundry at launch. OpenAI's own system card addendum, published September 29, 2026, confirms a knowledge cutoff of April 30, 2026 and treats the model as Critical in cybersecurity under the Preparedness Framework.

Ultrafast mode is the speed path. It runs up to 8x faster than Sol Standard at $12/M input and $60/M output, rolling out across the API, Codex, and ChatGPT Work — though in Codex and Work it is currently gated to the $500/month Pro tier.

Common Issues and Gotchas

Three recurring problems show up in the first two weeks of use.

Throughput. Users measured GPT-6.1 Sol xHigh at around 20.7 tok/s versus 53.8 tok/s for GPT-6 Sol xHigh — a real slowdown at the same price. OpenAI's Tibo acknowledged the heavy load and said capacity was expanded, with a global usage reset for paid accounts on October 2, 2026. If speed is critical, use Ultrafast.

Image input. At least one user reported GPT-6.1 Sol intermittently refusing to read images.

Source: @ItsmeAjayKV

Codex connectivity. A "Reconnecting… waiting for network" message has been reported in Codex as recently as October 11, 2026. Treat it as a client-side retry, not a billing failure.

Frequently Asked Questions

Does GPT-6.1 Sol have an API? Yes. GPT-6.1 Sol is available through OpenAI's API and through the kie.ai API under the model string gpt-6.1-sol, at $2 per million input tokens and $10 per million output tokens.

Can I use GPT-6.1 Sol for free? There is no free API tier for GPT-6.1 Sol. It is reachable inside Codex and ChatGPT Work for Plus, Pro, Business, Enterprise, and Edu subscribers, and through metered API billing on kie.ai and OpenAI.

What is the GPT-6.1 Sol model string? The OpenAI API model identifier is gpt-6.1-sol. On Amazon Bedrock and Box the name is namespaced differently, but the OpenAI-native string is gpt-6.1-sol.

How much does GPT-6.1 Sol cost per token? GPT-6.1 Sol costs $2 per million input tokens, $10 per million output tokens, and $0.10 per million cached input tokens — a 95% cache discount. Ultrafast mode is $12 input and $60 output per million.

Is GPT-6.1 Sol available in ChatGPT? At launch GPT-6.1 Sol ran only in ChatGPT Work and Codex, not the regular chat mode. OpenAI has since begun rolling GPT-6 Sol into the ChatGPT Chat tab for paid plans.

How do I enable Ultrafast mode on GPT-6.1 Sol? Ultrafast mode runs up to 8x faster than Sol Standard and is available in the API, Codex, and ChatGPT Work. In Codex and Work it is restricted to the $500/month Pro tier and eligible business and education accounts.

Why is GPT-6.1 Sol so slow? GPT-6.1 Sol launched under heavy load and measured around 20-30 tokens per second for many users. OpenAI added capacity and said speeds returned to expected levels; Ultrafast mode is the faster-throughput option.

What to Watch Next

Three signals matter for anyone building on GPT-6.1 Sol. First, whether Ultrafast opens beyond the $500 Pro tier — that decision sets the real cost of fast agent loops. Second, the arrival of GPT-6.1 Sol in regular ChatGPT chat, which OpenAI has flagged as "coming soon." Third, standard-tier throughput: the model trades speed for price, and OpenAI's 28-day improvement sprint is where any gains will land. For background on the model's benchmarks and positioning, see our earlier analysis, GPT-6.1 Sol Is a 1M-Token Agentic Workhorse.

Building similar agentic chat workflows? On kie.ai you can try GPT 6.1 Sol, GPT-6 Astra, and Claude Opus 5.5.