GPT-6.2 Sol Benchmark Results for Data Analysis

Nov 17, 2026: OpenAI's new everyday model, tested on real-world data analysis the day it launched.

At Plotly, we keep a close eye on model releases. Our agentic analytics products, Plotly Studio Desktop and Plotly Studio Embedded, and our open source libraries Plotly and Dash work with any model, and our customers look to us for recommendations.

Today OpenAI released GPT-6.2 Sol, which its announcement calls "our efficient, versatile model for everyday work", alongside GPT-6.2 Astra, "our new frontier model for your hardest projects" (our GPT-6.2 Astra results are here). OpenAI says that compared with GPT-6.1 Sol, GPT-6.2 Sol "is smarter, more efficient, and lets you complete more tasks on the same paid plan," and that it keeps GPT-6.1 Sol's prices: on OpenRouter, $2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens, with a context window of 1.05 million tokens and up to 128,000 output tokens. The weights are closed, and OpenAI's announcement doesn't say what GPT-6.2 Sol was built from or how it relates to GPT-6.1 Sol in training.

We ran it through our data analytics exam. GPT-6.2 Sol passed 40 of 43 questions (93%, an A) for $1.22 a full run, under half of GPT-6.1 Sol's $2.61. That's under 3 cents a question, or about $28 per 1,000 questions. Its median question took 10.5 seconds, which makes it the fastest A-grade model we've tested. It's also the third-cheapest, after Hy4 flash at $0.13 and Particle-1 at $0.35, and just under GPT-6.1 Terra at $1.31.

Data analytics benchmark cost-accuracy frontier with every model labelled and GPT-6.2 Sol circled at 93% for $1.22

GPT-6.2 Sol vs. GPT-6.1 Sol, GPT-6.1 Terra and Particle-1

GPT-6.2 Sol

GPT-6.1 Sol

GPT-6.1 Terra

Particle-1

Score

40 / 43

43 / 43

40 / 43

40 / 43

Analytics questions

20 / 20

20 / 20

20 / 20

20 / 20

Ambiguous questions

5 / 5

5 / 5

5 / 5

5 / 5

Impossible requests

5 / 5

5 / 5

5 / 5

5 / 5

Data quality

3 / 3

3 / 3

3 / 3

2 / 3

Statistics traps

5 / 5

5 / 5

4 / 5

5 / 5

Hidden needles

2 / 5

5 / 5

3 / 5

3 / 5

Open weights

No

No

No

No

Cost of one full run

$1.22

$2.61

$1.31

$0.35

Per person per year*

$71

$152

$76

$21

Median time per question

10.5 seconds

17.6 seconds

15.4 seconds

15.6 seconds

*1,000 people each asking 10 questions a day, 250 working days a year, at OpenRouter's prices when each model ran.

GPT-6.2 Sol ties three other models at 40: GPT-6.1 Terra, Particle-1 and Muse Spark 1.3, and 15 of the 53 models on our board score higher. GPT-6.1 Sol scored 43. Each model gets a single run on this benchmark, and in our repeat runs of other models scores moved by up to three points between runs, so a question or two either way is within normal run-to-run variation.

It isn't on the cost-accuracy frontier, because Particle-1 scores the same 40 for $0.35. It is on the speed-accuracy frontier, which is now Mercury 3, GPT-6.2 Sol, GPT-6.2 Astra and GPT-6.1 Sol: no model we've tested is both faster and more accurate.

Why it costs less

OpenAI kept GPT-6.1 Sol's token prices, so the whole saving comes from GPT-6.2 Sol doing less work to answer the same 43 questions.

GPT-6.1 Sol

GPT-6.2 Sol

Price per million input / output tokens

$2 / $10

$2 / $10

Price per million cached input tokens

$0.10

$0.10

Tool calls

258

158

Model calls

227

173

Input tokens

7.8 million

4.6 million

Output tokens

70,000

45,000

Input tokens read from cache

94%

97%

Cost of one full run

$2.61

$1.22

Totals for the full exam of 43 questions. The saving is across the board, not only on the hardest questions. On the 38 questions without a hidden pattern, GPT-6.2 Sol made about 2.9 tool calls per question against 4.3 for GPT-6.1 Sol, about a third fewer, and still answered every one correctly. On the five hidden-pattern questions it searched much less, running about 9 database queries per question against about 16, and that's where all three of its misses are.

What it got right

GPT-6.2 Sol was perfect on five of the six kinds of question. It got all 20 straightforward analytics questions, it asked about or stated its assumptions on all five ambiguous questions, and it turned down all five impossible requests instead of inventing an answer. It caught all three data quality problems we planted in the warehouse, and it avoided all five statistics traps, where the obvious calculation gives a misleading answer. GPT-6.1 Terra, at the same score, missed one of those traps, and Particle-1 missed one of the data quality problems.

It's efficient, too. It made about 4 model calls per question and wrote about 1,000 output tokens per question, and 97% of the input tokens it read came from the provider's cache. The run had no model errors.

What it got wrong

It missed three questions, and we checked each one by hand against the transcript and the answer key. All three were genuine misses, not grading problems.

All three were hidden patterns, the hardest part of our exam. Each one points toward something unusual in the data without naming it, and the model has to go looking. In one, GPT-6.2 Sol never isolated the pattern we planted. In another, it never named what was affected. In the third, it never looked at the part of the data where the effect was. Across the five hidden-pattern questions it ran about 9 queries each, against a median of 25 on our board.

Speed

GPT-6.2 Sol's median question took 10.5 seconds, and the whole exam about 13 minutes, against 17.6 seconds and about 23 minutes for GPT-6.1 Sol. Only Mercury 3 and gpt-oss-120b have a faster median, and they scored 35 and 27. Both GPT-6.2 models join the speed-accuracy frontier, shown here.

Accuracy vs. median time per question for every model on the board, with the speed-accuracy frontier

Which OpenAI model to choose

For data analysis, GPT-6.1 Sol is still the cheapest model we've tested with a perfect score, at $2.61 a run. GPT-6.2 Sol costs under half as much and takes about 40% less time per question. GPT-6.1 Terra scored the same 40 for slightly more, $1.31, with a median of 15.4 seconds (more in our GPT-6.1 Terra post). GPT-6.2 Astra scored 41 for $4.11 (GPT-6.2 Astra results).

OpenAI's models over time

We've tested 16 OpenAI models, from GPT-4o in November 2024 to the two released today, shown here with every other model in grey. GPT-4o passed 20 of 43 for $5.83; GPT-6.2 Sol passes twice as many for about a fifth of the cost. Speed has moved around more: GPT-4o was quick at 7 seconds a question but got fewer than half right, GPT-5 took 82, and the GPT-6.2 models, at 10.5 and 12.4 seconds, are the fastest OpenAI models we've tested that score an A.

Accuracy vs. cost of all 16 OpenAI models we've tested: GPT-6.2 Astra and GPT-6.2 Sol drawn solid, the earlier OpenAI models outlined and labelled, and all other models in grey
Accuracy vs. median time per question for all 16 OpenAI models we've tested: GPT-6.2 Astra and GPT-6.2 Sol drawn solid, the earlier OpenAI models outlined and labelled, and all other models in grey

The Test Harness

We run every model in Plotly Studio Embedded, a drop-in analytics agent that runs on Dash apps. This harness is tuned to run data analytics queries and make charts with multi-turn tool calling, with less emphasis on writing text. It's meant to put the data front and center.

Claude Sonnet 5.5 in Plotly Studio Embedded answering "Which is our best-performing wind farm?"

Frequently asked questions

How good is GPT-6.2 Sol at data analysis?

On PlotlyDataAnalyticsBench it passed 40 of 43 questions (93%, an A), tied with GPT-6.1 Terra, Particle-1 and Muse Spark 1.3. It was perfect on analytics questions, ambiguous questions, impossible requests, data quality problems and statistics traps, and found two of the five hidden patterns.

How much does GPT-6.2 Sol cost?

$2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens, the same as GPT-6.1 Sol. Our whole exam cost $1.22, under 3 cents a question, or about $71 per person per year at 10 questions a day.

Why is GPT-6.2 Sol cheaper than GPT-6.1 Sol if the prices are the same?

It does less work per question. On our exam it made 158 tool calls against 258 and read 4.6 million input tokens against 7.8 million, and more of what it read came from the cache.

Is GPT-6.2 Sol fast?

Yes. Its median question took 10.5 seconds, the fastest of any model we've tested that scores an A. GPT-6.1 Sol took 17.6.

Is GPT-6.2 Sol open source?

No. Its weights are closed. It's available through OpenAI's API, in ChatGPT Work and Codex, and on OpenRouter. For how open-weight models compare, see our open vs. closed models post.

Should I use GPT-6.2 Sol or GPT-6.2 Astra?

On our exam they're close: GPT-6.2 Sol scored 40 and GPT-6.2 Astra 41. GPT-6.2 Sol cost $1.22 against $4.11 and had a median of 10.5 seconds against 12.4 (GPT-6.2 Astra results).

About the benchmark

We run new models through the same exam as they're released, and we'll keep sharing what we find. We ran GPT-6.2 Sol through OpenRouter on the day it arrived there, with the same harness and grading as every other model.

To see all 53 models on one chart and in a full results table, along with cost, speed and how the exam works, see Introducing DataAnalyticsBench.

The exam is private and we're sharing it with researchers. We're keeping it private so that the answer key doesn't make its way into the training set. For the same reason, the data is fabricated: the warehouse is a synthetic wind-energy fleet, built to look like a real operator's data, so its answers don't exist anywhere else.

If you'd like access, reach out to Chris Parmer at Plotly.

bluesly
x
instagram
youtube
medium
fcaebook

Product

© 2026
Plotly. All rights reserved.
Cookie Preferences
aicpa/soc2