AI coding assistants · September 2026 Edition
GitHub Copilot vs Cursor
Eleven of twelve models named GitHub Copilot first on the direct prompt; two named Cursor. Both were named by all twelve models and GitHub Copilot carries 70 labels and Cursor 52, so the shares are not directly comparable.
Named in one category this edition.
Named in one category this edition.
First-choice share67%4%Of first choices across the direct, paraphrase, budget and scale prompts, 0 to 100.
Negative rate13%19%Negative labels as a share of the product's labels, 0 to 100.
Rank in category#1#3A position in a field of 8; printed, not drawn.
Labels7052A count; the two differ.
The two percentage rows are drawn on one 0 to 100 track, GitHub Copilot reading right to left. Rank and label count are printed, not drawn.GitHub Copilot vs Windsurf · Windsurf vs Cursor
Share is the count of first choices across the direct, paraphrase, budget and scale prompts over all twelve models, for a mid-market B2B company; rank is within the category; every quote names the model and the prompt it came from. Both figures come from the AI coding assistants page.
By framing
How many of the twelve models made each the first choice, per way of asking, and how many argued against it.
GitHub CopilotFirst choices, of twelve modelsCursor
Direct1121 against Cursor
Paraphrase1201 against Cursor
Comparative75
Budget-constrained601 against GitHub Copilot · 1 against Cursor
Scale-constrained30
Negative007 against GitHub Copilot · 7 against Cursor
Bars are first choices, 0 to 12 each sideModels that argued againstA model can name both, so the two sides of a row do not sum to twelve.
The direct prompt
The plain question, one answer per model, grouped by where GitHub Copilot and Cursor stood in it.
Both were the first choice
2 of 12 modelsThe answer named them together, and the judge labeled each a first choice.Mistral SmallCursor, GitHub Copilot alternatives: Claude Code
Llama 4 MaverickCursor, GitHub Copilot alternatives: Claude
GitHub Copilot first, Cursor an alternative
9 of 12 modelsCursor was named in the answer but not as the choice, or not at all.Claude Haiku 4.5Claude, GitHub Copilot
GPT-5.4 miniGitHub Copilot alternatives: Amazon Q Developer, Cursor
Perplexity SonarGitHub Copilot alternatives: Cursor, Sourcegraph Cody
Grok 4.1 FastGitHub Copilot alternatives: Amazon Q Developer, Cursor, Gemini Code Assist, Tabnine
DeepSeek V4 FlashGitHub Copilot alternatives: Amazon Q Developer, Cursor, Tabnine
Qwen 3.7 FlashGitHub Copilot alternatives: Cursor
Kimi K2GitHub Copilot alternatives: Tabnine, Windsurf
GLM 4.7 FlashXGitHub Copilot alternatives: Amazon Q Developer, Cursor, Tabnine
MiniMax M2.5GitHub Copilot alternatives: Claude, Claude Code, Cursor
Neither was the first choice, one was named
1 of 12 modelsThe answer put something else first and named one of the two as an alternative.Gemini 3.5 FlashAugment Code alternatives: GitHub Copilot, Tabnine
Bold names in an answer are the products the judge labeled a first choice; a model naming several gives each of them that label. The full answer text for every row is in the record.
By buyer segment
The same question asked on behalf of a different buyer. Each standing is computed within its segment and they are never added together. The figures above are the mid-market standing, which is the one the category orders by.
Small business
GitHub Copilot leads by fifty-six points.
GitHub Copilot65%#1 of 6Cursor9%#3 of 6Mid-marketThe figures above
GitHub Copilot leads by sixty-two points.
GitHub Copilot67%#1 of 8Cursor4%#3 of 8Enterprise
GitHub Copilot leads by ninety-five points.
GitHub Copilot95%#1 of 9Cursor0%#9 of 9What the models said about GitHub Copilot
Every negative label with a quote, up to three, then the highest-weighted positives, up to three. Six of eight in this category shown.
“You should generally avoid standard free-tier or public instances of major AI coding assistants (such as GitHub Copilot standard, Cursor, or Windsurf)” Qwen 3.7 Flash · negative prompt · hard negative
“Prompt injection vulnerabilities leading to arbitrary command execution (e.g., CVE-2025-53773)” GLM 4.7 FlashX · negative prompt · hard negative
“Avoid: Free tiers of Copilot/Windsurf for proprietary code” DeepSeek V4 Flash · negative prompt · hard negative
“GitHub Copilot (Best Value for Money) ... Codeium's free tier or GitHub Copilot's $10/month plan offer the best balance of cost and functionality.” Claude Haiku 4.5 · budget prompt · first choice
“I'd typically recommend starting with GitHub Copilot (if you're GitHub-centric) or Codeium (for cost efficiency and privacy options) as pilots” Claude Haiku 4.5 · paraphrase prompt · first choice
“Start with GitHub Copilot if you're unsure. It's the safest, most widely adopted choice with the strongest enterprise security posture.” Qwen 3.7 Flash · direct prompt · first choice
What the models said about Cursor
Every negative label with a quote, up to three, then the highest-weighted positives, up to three. Six of eight in this category shown.
“avoid standard free-tier or public instances of major AI coding assistants (such as GitHub Copilot standard, Cursor, or Windsurf)” Qwen 3.7 Flash · negative prompt · hard negative
“Over 30 vulnerabilities reported; 100% of tested AI IDEs were vulnerable to chained prompt injection and auto-approved tool calls” GLM 4.7 FlashX · negative prompt · hard negative
“Documented security incidents involving insufficient boundaries and autonomous actions” Kimi K2 · negative prompt · hard negative
“If you want a state-of-the-art, everyday editor that manages your entire workspace effortlessly: Go with Cursor.” Gemini 3.5 Flash · comparative prompt · first choice
“likely to be GitHub Copilot Business or Cursor Business, depending on the specific needs and budget” Llama 4 Maverick · direct prompt · first choice
“the best AI coding assistants are GitHub Copilot Business and Cursor Business” Mistral Small · direct prompt · first choice
Comparisons are drawn for the top three products in each category. The output is the models' output; nothing here is a recommendation by the index.