> ## Content Index
> Fetch the complete content index at: https://www.implicator.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# OpenAI Keeps GPT-6.1 Sol at $2 per Million Input Tokens and Halves Its Cached-Input Price
- URL: https://www.implicator.ai/openai-gpt-6-1-sol-cached-input-price/
- Published: 2026-09-29T20:06:23.000Z
- Updated: 2026-09-29T20:06:23.000Z
- Description: GPT-6.1 Sol keeps GPT-6 Sol's $2 and $10 token rates and halves cached input to $0.10. OpenAI's tests put it 2.1 points behind Astra on computer use at a seventh of the cost, while Claude Opus 5.5 still leads on some maximum-effort benchmarks.
- Author: Marcus Schuler
- Tags: AI News, #home-companion

[GPT-6.1 Sol](https://openai.com/index/introducing-gpt-6-1-sol/?ref=implicator.ai), released Tuesday, charges $0.10 per million cached input tokens, half GPT-6 Sol’s $0.20\. Standard input and output rates remain unchanged at $2 and $10 per million tokens. The saving applies to input the API serves from cache, the context agents resend on repeated calls.

The model arrived at OpenAI’s DevDay, one week after GPT-6 Sol. On OpenAI’s OSWorld 2.0 computer-use test, GPT-6.1 Sol trails GPT-6 Astra by 2.1 percentage points at $1.27 a task against $9.44\. Outside measurements from Artificial Analysis still rank it below the top Claude and Astra settings.

What Changed

- GPT-6.1 Sol charges $0.10 per million cached input tokens, half GPT-6 Sol's $0.20, while standard rates stay at $2 input and $10 output.
- On OpenAI's OSWorld 2.0 test it trails GPT-6 Astra by 2.1 points at $1.27 a task against $9.44.
- At maximum effort, Claude Opus 5.5 with fallbacks leads it on AutomationBench, 42.5% to 36.1%, and Artificial Analysis ranks it below top Claude and Astra settings.
- OpenAI treats the model as Critical in cybersecurity, and its coding misrepresentation rate rose to 1.50% from GPT-6 Sol's 1.30%.

AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.implicator.ai/about/).

## The price card

At launch, GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens, unchanged from [GPT-6 Sol’s rates](https://developers.openai.com/api/docs/models/gpt-6-sol?ref=implicator.ai).

GPT-6 Astra charges $10 for input, $1 for cached input and $50 for output per million tokens. GPT-6.1 Sol therefore costs one-fifth as much for standard input and output, and one-tenth as much for cached reads.

Claude Sonnet 5.5 has the same $2 input and $10 output rates per million tokens. Its cached input costs $0.20 per million tokens, twice GPT-6.1 Sol’s launch rate.

## What OpenAI’s charts show

Apart from the Artificial Analysis and Cognition measurements, the benchmark scores here come from OpenAI’s own runs, with competitor scores taken from public reports. OpenAI cautions that its research environment and API evaluations can produce different results from production ChatGPT.

On DeepSWE v1.1, which tests software engineering in real codebases, GPT-6.1 Sol scores 75.2% at high reasoning effort, against GPT-6 Astra’s 74.1% and GPT-6 Sol’s best score of 68.8%.

On OSWorld 2.0’s offline computer-use set, GPT-6.1 Sol scores 71.4% at maximum effort, against GPT-6 Astra’s 73.5%, with both figures measuring partial reward rather than the share of workflows completed. Average cost per task is $1.27 for GPT-6.1 Sol and $9.44 for GPT-6 Astra.

FREE AI BRIEFING · WEEKDAYS

Track what each AI model costs to run.

Get the AI stories shaping the day, with concise context from San Francisco. The briefing takes about five minutes and arrives at 4:45 a.m. Pacific, 7:45 a.m. Eastern.

Email address 

Keep me informed 

Check your inbox for the confirmation link.

Free. No hype. Unsubscribe anytime.

On GDP.pdf, which tests answers drawn from complex professional documents, GPT-6.1 Sol reaches 32.0% at high effort. Claude Opus 5.5 with fallbacks scores 28.8%.

On Terminal-Bench Science at maximum effort, GPT-6.1 Sol scores 57.0% at an average cost of $5.47 per task, against Claude Opus 5.5’s 63.3% for $23.21 and GPT-6 Astra’s 68.1% for $23.80\. OpenAI still recommends GPT-6 Astra for the hardest research tasks.

## Where the charts cut the other way

OpenAI’s claim that GPT-6.1 Sol beats Claude Opus 5.5 by 2.2 percentage points on AutomationBench uses medium-effort results: 31.7% against 29.5%.

At maximum effort, Claude Opus 5.5 with fallbacks leads on that business-workflow benchmark, scoring 42.5% against GPT-6.1 Sol’s 36.1%. GPT-6 Astra scores 41.4%.

GPT-6.1 Sol’s DeepSWE score also falls to 71.9% when reasoning effort rises from high to maximum.

[Artificial Analysis](https://artificialanalysis.ai/models/releases/gpt-6-1-sol?ref=implicator.ai) gives GPT-6.1 Sol at maximum effort a score of 52 on its Intelligence Index. Its list ranks that setting below Claude Opus 5.5, Claude Sonnet 5.5 and GPT-6 Astra settings.

On [Cognition’s FrontierCode 1.1](https://devin.ai/blog/gpt-6-1-sol?ref=implicator.ai), GPT-6.1 Sol scores 60.4%, close to GPT-6 Sol’s 60.7%. The new model costs $0.31 per task at medium effort, against $1.66 for GPT-6 Sol at maximum effort to reach its best score.

## The safety file

GPT-6.1 Sol has the first system-card addendum for a Sol model in this generation. Under OpenAI’s Preparedness Framework, it is treated as Critical in cybersecurity and High in biological and chemical capability. These are capability ratings, not rates of harm, and OpenAI’s framework defines Critical cyber capability as a model that can “identify and develop functional zero-day exploits of all severity levels in many hardened real-world critical systems without human intervention.” It uses GPT-6 Astra’s safeguards stack.

Know someone who'd find this useful? [✉️ Email it to a friend in one click](mailto:?subject=A%20newsletter%20I%20think%20you%27d%20like&body=This%20is%20one%20of%20maybe%20three%20newsletters%20I%20actually%20read.%20The%20rest%20just%20pile%20up%2C%20unread%2C%20judging%20me.%0A%0AAnd%20yes%2C%20this%20email%20mostly%20wrote%20itself%2C%20which%20is%20a%20little%20on%20the%20nose%20for%20an%20AI%20newsletter.%20Doesn%27t%20make%20it%20wrong.%20implicator.ai%20is%20good.%0A%0ASubscribe%20free%3A%20https%3A%2F%2Fwww.implicator.ai%2Fsubscribe%2F%3Futm%5Fsource%3Dnewsletter%26utm%5Fmedium%3Dforward%26utm%5Fcampaign%3Demail%5Fforward), or they can [subscribe free here](https://www.implicator.ai/subscribe/?utm%5Fsource=newsletter&utm%5Fmedium=forward&utm%5Fcampaign=forward%5Fto%5Fcolleague).

The [September 29 addendum](https://deploymentsafety.openai.com/gpt-6-1-sol?ref=implicator.ai) records unwanted persistence past explicit restrictions in 23.5% of GPT-6.1 Sol test runs, down from GPT-6 Sol’s 64.4%, but above GPT-6 Astra’s 17.4%. That evaluation ran without system controls designed to prevent circumvention.

Coding misrepresentation rises to 1.50% for GPT-6.1 Sol, against 1.30% for GPT-6 Sol and 0.51% for GPT-6 Astra. Failure to disclose an unavailable search tool falls to 2.08%, against GPT-6 Sol’s 4.92%.

OpenAI says these tasks were selected to provoke failures and do not reflect typical use.

## Where it runs

GPT-6.1 Sol is available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu customers, and through the API as `gpt-6.1-sol`. It is not yet available in regular Chat. Reasoning is always on, and tool calling requires the Responses API.

OpenAI says GPT-6.1 Sol Ultrafast will arrive in Codex “in the coming days.”

[GitHub Copilot’s rollout](https://github.blog/changelog/2026-09-29-gpt-6-1-sol-in-github-copilot/?ref=implicator.ai) covers Pro+, Max, Business and Enterprise accounts, with gradual availability and billing at provider list pricing. GitHub says early tests used fewer tokens and steps than earlier models, but published no reduction figures.

On Hacker News, commenter cmrdporcupine wrote: “6 Sol was worse than 5.6 Sol from my own experiences. Far worse. Will see if this remedies things.”

Frequently Asked Questions

What does GPT-6.1 Sol cost in the API?

$2 per million input tokens, $10 per million output tokens and $0.10 per million cached input tokens. Input and output rates are unchanged from GPT-6 Sol, while the cached-input rate is half GPT-6 Sol's $0.20.

How does it compare with GPT-6 Astra on price?

GPT-6 Astra charges $10 input, $1 cached input and $50 output per million tokens. GPT-6.1 Sol costs one-fifth as much for standard input and output and one-tenth as much for cached reads.

Does GPT-6.1 Sol beat Claude Opus 5.5?

On some OpenAI-run tests. It scores 32.0% on GDP.pdf against Opus 5.5's 28.8%. At maximum effort, Opus 5.5 with fallbacks leads on AutomationBench, 42.5% to 36.1%, and on Terminal-Bench Science, 63.3% to 57.0%.

Where can I use GPT-6.1 Sol?

In ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu customers, and through the API as gpt-6.1-sol. It is not yet in regular Chat. GitHub Copilot is rolling it out to Pro+, Max, Business and Enterprise accounts.

What does the safety addendum show?

Unwanted persistence past explicit restrictions in 23.5% of test runs, against 17.4% for GPT-6 Astra, and coding misrepresentation of 1.50%, against 1.30% for GPT-6 Sol. OpenAI says the tasks were selected to provoke failures.

AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.implicator.ai/about/).

[Anthropic cuts Opus 5.5 prices 20% and sends most cyber tasks to an older modelAnthropic released Claude Opus 5.5 on Tuesday, priced 20% below Opus 5, and it routes most cybersecurity requests to the older Opus 4.8\. The fallback happens automatically when Anthropic’s safeguards The Implicator![](https://www.implicator.ai/content/images/2026/09/2026-09-22-anthropic-opus-5-5-price-cut-cyber-routing.webp)](https://www.implicator.ai/anthropic-opus-5-5-price-cut-cyber-routing/)

[Meta Ties GPT-5.6 Sol at 42% Lower Cost; OpenAI Self-Rates Astra CriticalIMPLICATOR .ai Morning Briefing · From San Francisco Thursday, September 3, 2026 10 stops From San Francisco 1 The Editorial Good morning. Today's three picks wThe Implicator![](https://www.implicator.ai/content/images/2026/09/20260902-212415-nl_test_bay.webp)](https://www.implicator.ai/meta-muse-spark-ties-sol-openai-self-rates-astra/)

[Anthropic Ships Fable 5.1 at Same Price, Cuts Cache Reads 75% to $0.25 per MillionAnthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1 at Fable 5’s unchanged base price, while reducing the price of cache reads. The discount takes reused context to $0.25 per millThe Implicator![](https://www.implicator.ai/content/images/2026/09/2026-09-01-anthropic-fable-5-1-same-price-cache-reads-cut.webp)](https://www.implicator.ai/anthropic-fable-5-1-same-price-cache-reads-cut/)