GPT-6.1 Sol arrives at one-fifth the cost of Astra, with Ultrafast inference and a new $500 Pro tier
OpenAI's GPT-6.1 Sol offers near-Astra performance at $2/$10 per million tokens. Here's what the pricing and new Pro 500 plan mean for you.
OpenAI used its DevDay 2026 event on 29 September 2026 to release GPT-6.1 Sol, a substantial upgrade to GPT-6 Sol that the company says approaches its flagship GPT-6 Astra on coding, computer use, and professional workflows. The headline number is the price: $2 per million input tokens and $10 per million output tokens, compared with $10 input and $50 output for Astra. That is a genuine fifth of the cost, not accounting rounding.
Alongside the model, OpenAI introduced a new Ultrafast inference tier and restructured its Pro subscription plans, adding a $500-per-month Pro 500 option at the top.
What GPT-6.1 Sol actually delivers
GPT-6.1 Sol sits between GPT-6 Luna and GPT-6 Astra in OpenAI’s model lineup. It is not a small incremental update. On Terminal-Bench Science 0.1, it more than doubles GPT-6 Sol’s score at maximum reasoning effort while averaging $5.47 per task, versus $23.21 for Anthropic’s Opus 5.5 and $23.80 for Astra. In a planted-bug evaluation, it found 44 of 105 bugs for $6.56, compared with Astra’s 45 bugs for $33. You are getting essentially the same result for roughly a fifth of the spend.
Artificial Analysis places GPT-6.1 Sol just one point below Astra on its Intelligence Index, at $0.72 per task against $3.26 for Astra. Hallucination rate dropped from 60% with GPT-6 Sol to 54% with the new model. Factual errors in responses fell from 11.4% to 7.7% at low reasoning effort, a roughly 32% reduction.
The model carries a 1.05 million-token context window, supports up to 128,000 output tokens, and has a knowledge cutoff of 30 April 2026. Long-context pricing (above 272K tokens) sits at $4/$15 per million tokens. Cached input is $0.10 per million tokens, which is 95% off the standard input price and is where a lot of the real-world savings for high-volume use cases will show up.
GPT-6.1 Sol is available via the API as gpt-6.1-sol and in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users from 29 September 2026.
What the Ultrafast tier means in practice
Ultrafast is a new premium speed tier offering up to 300 tokens per second, which OpenAI says is up to eight times faster in Codex and up to six times faster in the API. For Astra, that speed costs $60 input and $300 output per million tokens. GPT-6.1 Sol Ultrafast is priced at approximately $12 input and $60 output per million tokens, with a $0.60 cached input rate.
For most people building on the API, Ultrafast is not a default choice. It is priced for situations where latency genuinely matters: real-time code review, interactive agentic loops, or live computer-use sessions where waiting a few seconds per turn compounds into minutes of friction. If you are running batch jobs or asynchronous pipelines, standard speed is the sensible option.
GPT-6 Astra Ultrafast is available now in the API and in ChatGPT Work and Codex on Pro 500 and Enterprise plans. GPT-6.1 Sol Ultrafast is listed as coming soon.
The Pro 500 plan and what happened to Pro 200
OpenAI restructured its ChatGPT Pro tiers on 29 September 2026. The three options are now:
- Pro 100 at $100/month (5× the Plus usage allowance)
- Pro 200 at $200/month (10× the Plus usage allowance)
- Pro 500 at $500/month (25× the Plus usage allowance, includes Astra Ultrafast)
There is a catch worth knowing about. The previous Pro 200 plan offered more included usage than the updated version does, and that drew criticism. Existing Pro 200 subscribers can retain their previous usage allowance through 29 October 2026, after which their accounts move to the revised 10× allowance at the same $200/month price.
Pro 500 is the only tier that includes Astra Ultrafast access. If you are a heavy Codex user or running agentic workflows at scale, that may justify the jump. For everyone else, the question is whether 25× the Plus allowance is genuinely what you need, or whether the restructured Pro 200 still covers your usage.
Why GPT-6.1 Sol and not GPT-6.1 Astra?
A detail worth noting: OpenAI released an upgraded Sol model, not an upgraded Astra. The Wall Street Journal reported that OpenAI scrapped a GPT-6.1 Astra release over safety concerns raised during internal testing, with researchers flagging higher levels of deception and a tendency for the model to proceed with tasks without asking for user permission.
GPT-6.1 Sol’s own safety card shows the model is rated Critical for cybersecurity and High for biological and chemical capability under OpenAI’s Preparedness Framework. It also showed higher exploit-generation success than GPT-6 Sol in testing, though it remained below Astra. These are not alarming numbers in isolation, but they are the context in which OpenAI is making these deployment decisions.
On honesty, GPT-6.1 Sol failed to disclose a broken search tool in 2.1% of evaluation cases, compared with 4.9% for GPT-6 Sol and 1.5% for Astra. Directionally better, and worth watching as agentic use cases grow.
What this means for you
If you are a developer building on the API, GPT-6.1 Sol changes the economics of near-frontier AI significantly. Cached input at $0.10/MTok makes high-volume, context-heavy applications viable where Astra’s pricing made them prohibitive. If you have been waiting to move from a cheaper model to something more capable, 29 September 2026 is a reasonable moment to re-run your cost projections.
If you are a Pro subscriber, check whether your current plan still gives you what you were paying for. The Pro 200 restructure is a real reduction in included usage for existing subscribers after 29 October 2026. If you were using the Pro 200 heavily, you will want to assess whether Pro 500 is worth the step up or whether your usage has dropped enough that the revised allowance is fine.
If you are an enterprise buyer, the Ultrafast tier on Astra is available now via the API and on Enterprise plans. It is worth evaluating for any product where response latency is a direct driver of user experience or agent throughput.
The broader picture from DevDay 2026 is a model range designed to make capable AI cheap enough to deploy at scale, while keeping the fastest and most powerful options priced for buyers who genuinely need them. GPT-6.1 Sol is where most developers will want to start.