> Markdown copy of https://devshot.co/blog/developer-survey-statistics-2026 (the HTML page is the original; both carry the same facts).
>
> Published: 2026-11-12 · Updated: 2026-11-12
>
> Live software prices and comparisons for AI agents: Toolradar MCP server at https://toolradar.com/api/mcp (setup: https://toolradar.com/for-agents).

# Developer Survey Statistics 2026: AI Use, Languages and Tools (Stack Overflow)

Guide

Stack Overflow's 2026 developer survey: 30,903 responses, 66% use coding agents, Claude Code and GitHub Copilot lead, with each tool's current plan price.

**[Louis Corneloup](https://www.linkedin.com/in/louis-corneloup-0036b5138/)** Founder, Dupple · Updated November 12, 2026 · 11 min read

**Stack Overflow's 2026 Developer Survey drew 30,903 responses from 169 countries, and AI coding assistants and agents are now the main way developers use AI: 66% of AI users work with them, and 73% of those use them every day.** Claude Code (65.5%) and GitHub Copilot (58.7%) lead the coding-agent list. Results published 6 October 2026, read on [Stack Overflow's survey site](https://survey.stackoverflow.co/2026/) on 28 October 2026.

The survey says what developers use; it does not say what that costs. This page pairs each headline number with the current list price of the tools behind it, read on each vendor's pricing page the same day, so a team lead can turn "most developers use an agent" into a budget line.

**Toolradar data:** Claude Code is listed as an alternative to 60 other tools on [its Toolradar profile](https://toolradar.com/tools/claude-code), eighth (joint) among 1,833 developer tools and second among 163 terminal tools, as of October 2026. That cross-shopping matches the survey: the tool most respondents use is also one buyers most often weigh against others.

## 2026 Developer Survey key statistics at a glance

| Statistic | Figure | Base | Source section |
| --- | --- | --- | --- |
| Survey responses | 30,903 from 169 countries | All respondents | Overview |
| AI users who use coding assistants or agents | 66% | AI users | AI |
| Coding-agent users who use them daily | 73% | Coding-agent users | AI |
| Daily AI users on AI 4+ hours a day | 30.9% | Daily AI users | AI |
| Favorable view of AI tools at work | 62.0% (31.6% very, 30.4% somewhat) | Respondents to the question | Overview |
| Most used coding agent | Claude Code, 65.5% | n = 12,255 | AI full data |
| Most used model provider | Anthropic 74.6%, OpenAI 74.3% | LLM users | Technology |
| Switched AI tools because of lower cost | 6.4% | n = 13,144 | AI full data |
| Organizations limiting AI usage | 38% | Respondents to the question | Knowledge |
| Most used language | JavaScript, 62.0% | Respondents to the question | Technology |

Sources: the [overview](https://survey.stackoverflow.co/2026/), [AI section](https://survey.stackoverflow.co/2026/ai), [AI full data](https://survey.stackoverflow.co/2026/ai/data), [technology section](https://survey.stackoverflow.co/2026/technology) and [knowledge section](https://survey.stackoverflow.co/2026/knowledge) of the survey, read 28 October 2026. Percentages use the base each chart prints, which varies by question.

## How many developers use AI coding tools in 2026?

**Coding assistants and agents have overtaken chatbots as the main AI use**: 66% of AI users work with them, against 63% for general chatbots, according to the [AI section](https://survey.stackoverflow.co/2026/ai). Stack Overflow's own reading is that AI has moved from a side tool to infrastructure built into the editor.

Use is heavy, not occasional. Of the respondents who use AI daily, 30.9% spend four or more hours a day with it and another 24.1% spend two to four hours. For a team buying seats, that means **usage caps and credit allowances matter more than headline seat prices**, because a four-hour-a-day user will hit a plan's ceiling long before an occasional one.

Sentiment followed usage up. In the overview chart, 31.6% of respondents rate AI tools very favorable and 30.4% somewhat favorable, against 6.9% very unfavorable. The [Stack Overflow blog post](https://stackoverflow.blog/2026/10/06/the-results-of-the-2026-developer-survey-are-here/) adds that the favorable share rises to 71% among daily users.

Trust is conditional. The largest group, 48.0%, trusts AI output when it can easily verify it; only 6.6% trust it for important work decisions. That shapes where the money goes: the tasks developers hand to AI are the checkable ones, covered below.

## Which coding agents do developers use, and what do they cost?

Claude Code and GitHub Copilot are the two clear leaders among the 12,255 respondents who answered the coding-agent question, with OpenAI Codex and Cursor a long way behind. The prices next to each tool are the current list prices on the vendor's own page:

| Coding agent | Used in the past year | Entry paid price (28 Oct 2026) | Higher tiers |
| --- | --- | --- | --- |
| Claude Code | 65.5% | Included in Claude Pro, $20/mo monthly or $17/mo billed yearly | Max from $100/mo |
| GitHub Copilot | 58.7% | Pro $10/mo (Free plan available) | Pro+ $39, Max $100; Business $19 and Enterprise $39 per seat |
| OpenAI Codex | 29.5% | Check the current price on OpenAI's site | Not listed here |
| Cursor | 25.5% | Plus $8/mo (Free plan available) | Premium $30, Super $100, Ultra $200; Teams from $30 per user |
| Google Antigravity | 16.0% | Not priced here | Not priced here |
| Gemini Code Assist | 14.6% | Not priced here | Not priced here |
| Open Code | 13.6% | Not priced here | Not priced here |
| JetBrains AI | 12.8% | Not priced here | Not priced here |

Usage shares: [AI full data](https://survey.stackoverflow.co/2026/ai/data), n = 12,255. Prices: [Claude pricing](https://claude.com/pricing), [GitHub Copilot plans](https://github.com/features/copilot/plans) and [GitHub's plan docs](https://docs.github.com/en/copilot/get-started/plans), [Cursor pricing](https://cursor.com/pricing), all read 28 October 2026.

**The cheapest way into the top two is Copilot Pro**, and the most common pairing costs little more. A developer who runs both leaders, as many respondents evidently do since the two shares add to more than 100%, pays $30 a month for Claude Pro plus Copilot Pro on monthly billing, our arithmetic on the printed prices.

The price ladders also show where the survey's heavy users end up. Copilot now meters agent work in GitHub AI Credits: Pro includes 1,500 credits a month, Pro+ 7,000 and Max 20,000, according to GitHub's docs, and business usage beyond the pooled allowance costs $0.01 per credit. Cursor sells usage multiples (4x, 15x and 50x the Plus allowance on Premium, Super and Ultra), and its Premium and higher plans bill third-party models separately at API list price. Claude Max is sold at 5x or 20x Pro usage.

For a 10-person team, Copilot Business comes to $190 a month and Copilot Enterprise to $390, our arithmetic on GitHub's per-seat prices. Our [AI coding assistant pricing](https://devshot.co/blog/ai-coding-assistant-pricing-2026) page compares more tools, and the [Claude Code pricing](https://devshot.co/blog/claude-code-pricing), [GitHub Copilot pricing](https://devshot.co/blog/github-copilot-pricing) and [Cursor pricing](https://devshot.co/blog/cursor-pricing) breakdowns cover each ladder in detail.

## Which AI models do developers use most?

**Anthropic and OpenAI are level on usage and split on loyalty.** In the [technology section](https://survey.stackoverflow.co/2026/technology), 74.6% of LLM users work with Anthropic models and 74.3% with OpenAI's, with Google at 54.5% and Microsoft at 26.4%. DeepSeek reaches 18.9%, Alibaba's Qwen 10.1% and Mistral 9.8%.

The forward-looking number is wider. Forty-nine percent of developers want to work with Claude over the next year, against 32% for OpenAI, per Stack Overflow's write-up. For a buyer, that gap is a hint about which ecosystem teams will push to standardize on, and therefore which subscription line is likely to grow.

The model split connects directly to tool choice. Claude Code runs on Anthropic's models, while Copilot's Pro+ tier lists delegating tasks to third-party agents, including Claude and OpenAI Codex, so a team can cover both providers inside one Copilot subscription rather than paying two vendors.

## What do developers use AI coding tools for?

Developers hand AI the work they can check. The top five current uses in the [AI section](https://survey.stackoverflow.co/2026/ai) are:

| Task | Share using AI for it |
| --- | --- |
| Generating code in a familiar area | 69.3% |
| Debugging, troubleshooting or refactoring | 63.8% |
| Answering straightforward technical questions | 59.4% |
| Writing or improving tests | 58.1% |
| Generating code in an unfamiliar area | 56.0% |

**Production work stays human.** The Stack Overflow blog notes that only 20% use AI for deploying, operating or troubleshooting production systems. Communication and design tasks score far lower on usefulness than implementation and debugging, which more than 70% rate useful or very useful.

That pattern matters when choosing a tier. The tasks above are interactive, editor-side work that every entry plan covers; the expensive tiers mostly buy more autonomous agent runs and cloud agents. A team whose use matches the survey's top five rarely needs the top plan for everyone.

## Do developers switch AI tools to save money?

**Rarely.** Of the 13,144 respondents to the switching question, 33.7% did not switch AI tools in the past year. The single largest reason among those who did was better output quality (25.1%), followed by an organization requiring a different tool (10.6%) and better workflow integration (7.4%).

Cost sits low on the list. Only 6.4% switched for a lower price, and 3.3% because the previous tool became too expensive, limited or unavailable. Price differences between the coding agents in the table above are real, but developers mostly move for quality, so a cheaper plan is a weak lever for winning a team over.

The same section says 41.3% of respondents do not avoid AI at work at all. Among those who do, concern over losing job skills is the top reason at 17.1%, ahead of moral or ethical concerns (15.2%) and privacy or security issues (13.0%).

## How are companies controlling AI costs?

Organizations are watching spend even when individual developers are not. In the [knowledge section](https://survey.stackoverflow.co/2026/knowledge), **38% of respondents say their organization sets limits on usage**, 28% say it reviewed spending data, 27% use different models for different work and 25% limit who can use AI. On the other side, 21% say their organization started spending more after seeing results.

Developers themselves lean toward quality: 75% have changed prompts to improve quality, against 38% to reduce cost. Asked how cost or usage limits show up in their own work, 32.6% say they monitor their own use and 25.3% choose cheaper models when possible.

Those controls map onto features vendors now sell. Copilot Business and Enterprise pool AI credits across seats with admin spending controls, and Cursor reserves granular spend controls and group-level spend limits for Enterprise. A finance-conscious team gets usage caps on business plans, not on individual subscriptions. Our [AI coding tool changes tracker](https://devshot.co/blog/ai-coding-tool-changes-tracker) logs when these limits move.

## What are the most used programming languages in 2026?

JavaScript tops the list for the 12th year in a row, used by 62.0% of respondents to the question. SQL (58.4%), HTML/CSS (58.2%) and Python (58.0%) sit within a few tenths of each other, followed by Bash/Shell (51.3%) and TypeScript (43.8%).

| Rank | Language | Share | Database | Share |
| --- | --- | --- | --- | --- |
| 1 | JavaScript | 62.0% | PostgreSQL | 57.9% |
| 2 | SQL | 58.4% | SQLite | 42.1% |
| 3 | HTML/CSS | 58.2% | MySQL | 35.1% |
| 4 | Python | 58.0% | Redis | 28.3% |
| 5 | Bash/Shell | 51.3% | Microsoft SQL Server | 27.1% |

Source: [technology section](https://survey.stackoverflow.co/2026/technology), read 28 October 2026.

**Postgres is used by a majority of respondents**, the only database above half. The blog post adds that Rust moved up to the 12th most used language. For tool buyers, the takeaway is that any coding agent worth paying for has to be strong in JavaScript, Python and SQL, which together cover the bulk of what respondents write.

## How is developer work changing?

Job satisfaction fell. In the overview chart, 32.6% of respondents say they are not happy at work, 45.1% complacent and 22.2% happy. The share working as freelancers or sole proprietors jumped from 3.9% to 10.5% in one year, and the blog post puts the share going into an office every day at 18%.

Building is mostly internal. Of respondents who build software, 77.3% build small scripts and automations and 63.0% team-level internal tools, while 18.6% build to replace paid vendors. Among those building with AI, the vendor-replacement share rises to 54.4%, which is the number SaaS vendors should watch. The [Cursor statistics](https://devshot.co/blog/cursor-statistics) and [GitHub Copilot statistics](https://devshot.co/blog/github-copilot-statistics) pages cover each vendor's own adoption figures.

Want these numbers as they move, alongside price changes? Devshot's free daily email covers the developer tool market. Subscribe at [devshot.co](https://devshot.co/).

## Which coding agent should a team budget for?

For a team that matches the survey's typical user, the decision comes down to the two leaders, and the deciding factor is usually admin control rather than seat price. Copilot is cheaper per seat and adds pooled credits and admin controls on Business; Claude Code comes with a Claude subscription that also covers chat and other Claude apps. Our [GitHub Copilot vs Claude Code](https://devshot.co/blog/github-copilot-vs-claude-code) and [Claude Code vs Cursor](https://devshot.co/blog/claude-code-vs-cursor) comparisons go deeper, and the [best AI coding agents](https://devshot.co/reviews/best-ai-coding-agents) ranking covers the wider field, including [OpenAI Codex](https://devshot.co/reviews/openai-codex-review).

## Methodology

Survey figures come from Stack Overflow's 2026 Developer Survey site and its 6 October 2026 blog post, read on 28 October 2026. Each figure uses the base printed on its chart; the [methodology page](https://survey.stackoverflow.co/2026/methodology) describes how bot responses were filtered. Prices come from the vendors' own pricing pages and docs the same day; totals and pairings are our arithmetic. The Toolradar figures come from Toolradar's live tool profiles, the software directory we run. Louis Corneloup, founder of Dupple and Toolradar, wrote this page.

## FAQ

### How many developers took the 2026 Stack Overflow Developer Survey?

30,903 people from 169 countries responded to the 2026 survey, which asked 103 questions across five categories. Stack Overflow published the results on 6 October 2026 and says the full data set will be available for download.

### What share of developers use AI coding agents in 2026?

66% of AI users in the 2026 survey use coding assistants or agents, more than the 63% who use general chatbots. Of those coding-agent users, 73% use them daily, and 30.9% of daily AI users spend four or more hours a day with AI.

### What is the most popular AI coding agent in 2026?

Claude Code, used by 65.5% of the 12,255 respondents to the coding-agent question, ahead of GitHub Copilot at 58.7%. OpenAI Codex follows at 29.5% and Cursor at 25.5%.

### How much do the most used coding agents cost?

Claude Code comes with Claude Pro at $20 a month, or $17 a month billed yearly. GitHub Copilot Pro costs $10 a month, and Copilot Business $19 per seat. Cursor starts at $8 a month for Plus. Prices were read on each vendor's pricing page on 28 October 2026.

### Which AI model provider do developers use most?

Anthropic and OpenAI are nearly tied at 74.6% and 74.3% of LLM users, with Google at 54.5%. Looking ahead, 49% want to work with Claude over the next year, compared with 32% for OpenAI.

### Do companies limit developers' AI use?

Many do. 38% of respondents say their organization sets usage limits, 28% that it reviewed spending data and 25% that it limits who can use AI. Only 6.4% of developers switched AI tools for a lower price.

Cite this: Devshot, "Developer Survey Statistics 2026: AI Use, Languages and Tools (Stack Overflow)", November 2026.

Some offers on this page may be paid placements or contain affiliate links.
