Which AI Is Best for Engineering in 2026?

Engineers now use AI chatbots to check a beam calculation, look up a code requirement or write a test report, and some firms have begun testing AI on design work. Researchers have started to measure this as well, including the EngDesign benchmark presented at NeurIPS 2025, which runs AI-generated designs through engineering simulations. To find out which AI is best for engineering, our research team ran a head-to-head study from September 29 to October 16, 2026.

We gave the same 400 engineering problems to seven AI platforms, producing 2,800 answers. The problems covered eight tasks, with 50 problems each, across mechanical, civil and structural, electrical, chemical and aerospace engineering, and many included a drawing, schematic or photo of a failed part. Every platform was tested on its best available model. A panel of six licensed Professional Engineers (PEs) scored each answer from 0 to 100 without knowing which platform wrote it. We also checked 1,860 numerical results against hand calculations, checked every code or standard each platform cited against the current edition, and tracked how each platform handled 120 safety-critical problems, such as sizing a beam, a pressure vessel or a circuit breaker.

Best AI for Engineering Overall

A platform’s overall score is the average of its scores across the eight tasks in our study, with each task weighted equally. In the table below, we rank the seven platforms from highest to lowest.

Engineering Scores by AI Platform (September to October 2026)

Rank	AI Platform	Model Tested	Overall Score	Strongest Task
1	ChatGPT	GPT-6 Sol	90.0	Engineering Calculations
2	Claude	Claude Opus 5.5	89.8	Design Review
3	Google Gemini	Gemini 4 Argon	88.7	Engineering Calculations
4	Microsoft Copilot	GPT-6 Astra	85.3	CAD and Scripting
5	Grok	Grok 4.6	83.8	Engineering Calculations
6	DeepSeek	DeepSeek V4-Pro	82.4	Engineering Calculations
7	Meta AI	Muse Spark	80.0	Reading Drawings and Schematics

Our researchers took three main points from this data:

• We found that ChatGPT’s lead comes from its numerical work. It finished first or second in seven of the eight tasks and had the highest share of correct calculations in the study (95.8%).
• Claude finished 0.2 points behind ChatGPT and was strongest on judgment-based work, winning design review, failure analysis and technical reports (91.4). Its weakest task was reading drawings and schematics (86.9).
• DeepSeek scored 88.3 on engineering calculations, higher than Microsoft Copilot and Grok, but 76.4 on reading drawings and schematics, the lowest score for that task, because it often misread dimensions and symbols in images.

Best AI for Each Engineering Task

We split the problems evenly across eight tasks, with 50 problems per task. The first table below shows the winner and runner-up for each task. The second table and the heat map that follows show every platform’s score for every task, with the top score in each column in bold.

Best AI Platform for Each Engineering Task (September to October 2026)

Engineering Task	Best Platform	Score	Runner-Up	Score
Engineering Calculations	ChatGPT	92.4	Google Gemini	90.6
Codes and Standards	Google Gemini	90.4	Claude	89.1
Reading Drawings and Schematics	ChatGPT	91.8	Google Gemini	90.3
Materials Selection	Google Gemini	90.1	ChatGPT	88.9
Design Review	Claude	91.6	ChatGPT	89.4
Failure Analysis	Claude	91.3	ChatGPT	88.7
CAD and Scripting	ChatGPT	91.2	Claude	89.8
Technical Reports	Claude	91.4	ChatGPT	88.9

Engineering Scores by Task and AI Platform (September to October 2026)

AI Platform	Calculations	Codes	Drawings	Materials	Design Review	Failure Analysis	CAD and Scripting	Reports
ChatGPT	92.4	88.6	91.8	88.9	89.4	88.7	91.2	88.9
Claude	90.1	89.1	86.9	88.2	91.6	91.3	89.8	91.4
Google Gemini	90.6	90.4	90.3	90.1	87.2	86.8	87.9	86.4
Microsoft Copilot	86.4	85.1	85.9	84.6	84.2	83.8	86.7	86.1
Grok	87.1	81.3	84.2	83.1	83.6	82.9	85.8	82.4
DeepSeek	88.3	79.6	76.4	82.8	82.1	81.7	86.2	81.9
Meta AI	80.2	78.4	82.6	80.9	78.8	78.1	79.3	81.6

Here is what our team took from the task-level results:

• We found that ChatGPT and Claude won three tasks each and Google Gemini won two. No other platform finished first in any task.
• Reading drawings and schematics produced the widest gap of any task, 15.4 points between ChatGPT (91.8) and DeepSeek (76.4). The best platforms read dimensions, tolerances and wiring symbols from a photo of a drawing with few mistakes.
• Google Gemini led codes and standards (90.4) and materials selection (90.1). Judges said it was the most likely to quote the correct section of a building, electrical or pressure vessel code and to note which edition it used.

Accuracy and Safety: Calculations, Codes and Engineer Review

An engineering answer is only useful if the numbers and references are right. We checked 1,860 numerical results against hand calculations, counted unit errors such as mixing pounds and newtons, and checked every code or standard cited in the 640 answers that relied on one. For the 120 safety-critical problems, we counted any error that would lead to an unsafe design, and how often each platform said a licensed PE should review the result before it was used.

Accuracy and Safety Results by AI Platform (September to October 2026)

AI Platform	Calculations Correct	Errors on Safety-Critical Problems	Code or Standard Cited Incorrectly	Unit Errors per 100 Calculations	Recommended PE Review
ChatGPT	95.8%	2.1%	6.4%	0.8	84.6%
Claude	94.9%	1.6%	5.1%	0.6	93.2%
Google Gemini	94.6%	2.4%	3.8%	0.9	87.9%
Microsoft Copilot	92.7%	3.6%	7.9%	1.4	81.3%
Grok	93.1%	3.9%	11.6%	1.7	68.4%
DeepSeek	93.8%	3.2%	14.2%	1.2	62.7%
Meta AI	88.4%	5.8%	13.7%	2.6	71.9%

Our researchers drew three conclusions from the accuracy data:

• We found that ChatGPT got the most calculations right (95.8%), while Claude made the fewest errors on safety-critical problems (1.6%, or two of 120). Meta AI made the most, 5.8%.
• Incorrect code citations were the most common error for Grok (11.6%) and DeepSeek (14.2%), which often cited an older edition of a standard or a section number that does not exist.
• Claude said a licensed engineer should review the work in 93.2% of safety-critical answers, compared with 62.7% for DeepSeek. Earlier research points to the same need for review: in a University of Iowa study, GPT-4 answered 75.4% of questions correctly on a practice Fundamentals of Engineering exam in environmental engineering after its prompts were refined.

Best AI for Each Engineering Discipline

Engineers in different fields get different results. The table below shows each platform’s average score on the problems from each discipline, with the top score in each column in bold.

Engineering Scores by Discipline and AI Platform (September to October 2026)

AI Platform	Mechanical	Civil and Structural	Electrical	Chemical	Aerospace
ChatGPT	90.6	89.1	91.2	89.4	90.3
Claude	89.9	90.4	88.6	90.1	89.7
Google Gemini	88.4	89.6	88.1	88.9	87.8
Microsoft Copilot	85.6	85.2	86.4	84.9	84.8
Grok	84.4	82.7	84.9	83.1	83.6
DeepSeek	83.1	81.4	82.9	82.6	81.8
Meta AI	80.6	79.4	80.8	79.6	79.3

Here is what our team took from the discipline results:

• We found that ChatGPT scored highest for mechanical (90.6), electrical (91.2) and aerospace engineering (90.3), the fields with the most calculation-heavy problems.
• Claude led civil and structural (90.4) and chemical engineering (90.1), where more of the problems asked for design review and failure analysis.
• Google Gemini’s best discipline was civil and structural engineering (89.6), helped by its accuracy on building codes.

Best AI Models for Engineering, Including Free Options

Most platforms give free users a different model than paying subscribers. We tested 11 models across the seven platforms, including each platform’s free option. The table below ranks every model we tested.

Engineering Scores by Model (September to October 2026)

Rank	Model	Platform	Available On	Score
1	GPT-6 Sol	ChatGPT	Paid	90.0
2	Claude Opus 5.5	Claude	Paid	89.8
3	Claude Sonnet 5.5	Claude	Free and paid	88.9
4	Gemini 4 Argon	Google Gemini	Paid	88.7
5	Gemini 3.8 Flash	Google Gemini	Free and paid	86.9
6	GPT-6 Luna	ChatGPT	Free and paid	86.8
7	GPT-6 Astra	Microsoft Copilot	Free and paid	85.3
8	Grok 4.6	Grok	Free and paid	83.8
9	DeepSeek V4-Pro	DeepSeek	Free	82.4
10	DeepSeek V4.1 Flash	DeepSeek	Free	80.9
11	Muse Spark	Meta AI	Free	80.0

Our team took three main points from the model-level results:

• We found that Claude Sonnet 5.5 is the best free AI for engineering, scoring 88.9, behind only GPT-6 Sol (90.0) and Claude Opus 5.5 (89.8).
• Paying for ChatGPT added the most of any platform: GPT-6 Sol scored 3.2 points higher than the free GPT-6 Luna, mostly from better calculations and drawing reading.
• DeepSeek, which is free for everyone, scored 88.3 on engineering calculations with V4-Pro, a strong free option for numerical work that does not involve drawings.

Requesting a Copy of This Report

If you’d like to request a PDF copy of this report or learn more about our agency, you can reach out here.