GPT-6.1 Sol is OpenAI's new mid-tier model, launched at DevDay on September 29, 2026. OpenAI says it nearly matches GPT-6 Astra on agentic coding, computer use and professional work, at one-fifth of Astra's token price. For most production workloads, that makes it the model to benchmark against first.
Here is what GPT-6.1 Sol costs, how it performs, and when you should still pay for Astra. (New to the event? Start with our OpenAI DevDay 2026 recap.)
Where GPT-6.1 Sol fits in the GPT-6 family
OpenAI's GPT-6 line now has three tiers. GPT-6 Astra arrived on September 3, 2026, followed by GPT-6 Sol and GPT-6 Luna on September 22. Less than a week later, Sol received its 6.1 upgrade.
- GPT-6 Astra: the most capable model, for the hardest reasoning and research problems.
- GPT-6.1 Sol: the balanced option, close to Astra on most practical work.
- GPT-6 Luna: the fast, cheap model for high-volume everyday tasks.
GPT-6.1 Sol pricing
Standard API prices, per million tokens (OpenAI):
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6 Astra | $10.00 | $1.00 | $50.00 |
| GPT-6.1 Sol | $2.00 | $0.10 | $10.00 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 |
The standout number is cached input at $0.10 per million tokens. That is 95% below Sol's standard input price and half of GPT-6 Sol's cached rate. Agents that resend the same system prompt, tool definitions or codebase context on every step benefit the most.
A quick cost example
Imagine an agent that makes 10,000 calls a day. Each call sends 20,000 input tokens, 15,000 of them cached, and returns 1,000 output tokens.
- GPT-6.1 Sol: about $0.0215 per call, or roughly $215 a day.
- GPT-6 Astra: about $0.115 per call, or roughly $1,150 a day.
That is around 81% cheaper for the same workload shape. Your real numbers depend on reasoning effort, since higher effort produces more output tokens.
GPT-6.1 Sol benchmarks
OpenAI published results across coding, documents, workflows, computer use and science. All figures below are OpenAI's own and compare against GPT-6 Sol, GPT-6 Astra and Anthropic's Claude Opus 5.5.
| Benchmark | What it tests | GPT-6.1 Sol result |
|---|---|---|
| DeepSWE v1.1 | Long-horizon software engineering in real codebases | Matches Astra at about one-fifth of the cost; beats GPT-6 Sol's best by 6.4 points at lower effort |
| GDP.pdf | Answering professional questions from complex PDFs | Beats Opus 5.5 at under half the cost per task; approaches Astra at about one-fifth the cost |
| AutomationBench 1.0.6 | Multi-step business workflows across 47 tools | 2.2 points above Opus 5.5 at medium effort for about a third of the cost; up 4.8 points on GPT-6 Sol |
| OSWorld 2.0 (offline) | Long computer-use workflows | 7 points above GPT-6 Sol at max effort; within 2.1 points of Astra at about one-seventh the cost |
| Terminal-Bench Science 0.1 | Data analysis, simulation and theorem proving | More than double GPT-6 Sol's score; $5.47 per task vs $23.21 (Opus 5.5) and $23.80 (Astra) |
Astra still wins on the hardest science work, topping Terminal-Bench Science at 68.1%. OpenAI itself recommends Astra for the most difficult research tasks.
Fewer factual errors
On a deliberately hard set of prompts where users had flagged mistakes from earlier models, GPT-6.1 Sol at low reasoning effort cut the share of answers with a factual error from 11.4% to 7.7%. That is a reduction of roughly 32% versus GPT-6 Sol, and within 1.9 points of Astra across settings.
Safety and alignment
OpenAI says GPT-6.1 Sol is more transparent about its limits and better at respecting explicit restrictions during agentic tasks. In one stress test, it failed to tell users about a broken search tool 2.1% of the time, compared with 4.9% for GPT-6 Sol and 1.5% for Astra. OpenAI also reports no attempts to bypass its automated safety reviewer. Full details are in the GPT-6.1 Sol system card addendum.
As with any vendor benchmark, run your own evals on your own tasks before switching production traffic.
How to use GPT-6.1 Sol
In the API, call the model ID gpt-6.1-sol. A minimal Python example:
from openai import OpenAI
client = OpenAI()
response = client.responses.create(
model="gpt-6.1-sol",
reasoning={"effort": "medium"},
input="Refactor this Express route to use async/await and add input validation.",
)
print(response.output_text)Check the OpenAI API docs for the current parameter names and supported effort levels.
In ChatGPT, GPT-6.1 Sol is available now in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. It is not yet available in standard Chat.
Ultrafast for GPT-6.1 Sol is coming in the coming days, with up to 8x faster generation in Codex. BGR reports it will reach up to 350 tokens per second, and that Ultrafast costs six times the standard API price (BGR).
Should you switch to GPT-6.1 Sol?
- Switch from GPT-6 Sol: almost certainly. It is better across the board and cached input is half the price.
- Switch from GPT-6 Astra: test it for coding agents, computer use, document Q&A and business automation. The savings are large if quality holds.
- Stay on Astra: for frontier research, hard math and tasks where a few points of accuracy justify 5x the cost.
- Use Luna instead: for classification, extraction and simple chat at massive volume.
A common pattern is to route by difficulty: Luna for triage, GPT-6.1 Sol as the default worker, and Astra only for escalations. For more on building that kind of pipeline, see DevDay 2026 for Developers.
FAQ
What is GPT-6.1 Sol? An upgraded version of OpenAI's GPT-6 Sol model, released September 29, 2026, tuned for agentic coding, computer use and professional work.
How much does GPT-6.1 Sol cost? $2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens.
Is GPT-6.1 Sol better than GPT-6 Astra? No, but it comes close. OpenAI says it matches or nearly matches Astra on several coding and workflow benchmarks at about one-fifth of the price.
What is the GPT-6.1 Sol API model name? gpt-6.1-sol.
Is GPT-6.1 Sol available in ChatGPT? Yes, in ChatGPT Work and Codex for paid plans, but not yet in standard Chat.
Sources
- Introducing GPT-6.1 Sol, OpenAI
- DevDay 2026 Recap, OpenAI
- Everything OpenAI Announced at DevDay 2026, BGR
Also Read: OpenAI DevDay 2026: All 20+ Announcements Explained
