AI coding assistants
Asked as “AI coding assistant”, and as “AI pair programming tool”, on behalf of a mid-market B2B company. 51 first choices recorded across the direct, paraphrase, budget and scale prompts, fourteen models each.
Standing · first-choice share
75%
Clear leader
75GitHub Copilot14Windsurf04Amazon Q Developer08others
75% of first choices, clear leader.
Since September 2026▲+8Since September 2026: 67% → 75%, +8 points. Inside the 11-point floor: within noise. Read over the models both editions asked.GitHub Copilot held the lead, +8 points on 67%, inside the 11-point floor.
By buyer segment
The same question asked on behalf of a different buyer. Each standing is computed within its segment; they sit side by side and are never added together.
Small business
GitHub Copilot63%Clear leader▼−4Since September 2026: 65% → 61%, −4 points. Inside the 11-point floor: within noise. Read over the models both editions asked.Standing →Mid-marketGitHub Copilot75%Clear leader▲+8Since September 2026: 67% → 75%, +8 points. Inside the 11-point floor: within noise. Read over the models both editions asked.This pageEnterpriseGitHub Copilot82%Clear leader▼−13Since September 2026: 95% → 82%, −13 points. Past the 11-point floor: movement. Read over the models both editions asked.Standing →
GitHub Copilot63%Clear leader▼−4Since September 2026: 65% → 61%, −4 points. Inside the 11-point floor: within noise. Read over the models both editions asked.Standing →Mid-marketGitHub Copilot75%Clear leader▲+8Since September 2026: 67% → 75%, +8 points. Inside the 11-point floor: within noise. Read over the models both editions asked.This pageEnterpriseGitHub Copilot82%Clear leader▼−13Since September 2026: 95% → 82%, −13 points. Past the 11-point floor: movement. Read over the models both editions asked.Standing →The standing
Share is the count of first choices across the direct, paraphrase, budget and scale prompts, over all fourteen models, for a mid-market B2B company. Ordered by share.
ProductFirst-choice shareNegative rateLabelsQuadrantSince September 2026
01GitHub Copilot75%11%73endorsed leader▲+8Since September 2026: 67% → 75%, +8 points. Inside the 11-point floor: within noise. Read over the models both editions asked.67% → 75%
GitHub Copilot75%11%73endorsed leader▲+8Since September 2026: 67% → 75%, +8 points. Inside the 11-point floor: within noise. Read over the models both editions asked.67% → 75%02
Windsurf14%9%34accepted challenger▲+3Since September 2026: 10% → 14%, +3 points. Inside the 11-point floor: within noise. Read over the models both editions asked.10% → 14%
Windsurf14%9%34accepted challenger▲+3Since September 2026: 10% → 14%, +3 points. Inside the 11-point floor: within noise. Read over the models both editions asked.10% → 14%03
Cursor4%19%57accepted challenger=heldSince September 2026: 4% → 5%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.4% → 5%
Cursor4%19%57accepted challenger=heldSince September 2026: 4% → 5%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.4% → 5%04
Amazon Q Developer4%16%38accepted challenger=heldSince September 2026: 2% → 2%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.2% → 2%
Amazon Q Developer4%16%38accepted challenger=heldSince September 2026: 2% → 2%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.2% → 2%05
Tabnine2%0%33accepted challenger▲+2Since September 2026: 0% → 2%, +2 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 2%
Tabnine2%0%33accepted challenger▲+2Since September 2026: 0% → 2%, +2 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 2%Show the four products at 0%, ordered by negative rate
06
Claude Code Lab in the set0%14%28accepted challenger=heldSince September 2026: 0% → 0%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 0%
Claude Code Lab in the set0%14%28accepted challenger=heldSince September 2026: 0% → 0%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 0%08
Continue0%10%10accepted challenger=heldSince September 2026: 0% → 0%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 0%
Continue0%10%10accepted challenger=heldSince September 2026: 0% → 0%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 0%09
Devin0%10%10accepted challenger=heldSince September 2026: 0% → 0%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 0%
Devin0%10%10accepted challenger=heldSince September 2026: 0% → 0%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 0%07
Sourcegraph Cody0%0%10accepted challenger=heldSince September 2026: 0% → 0%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 0%
Sourcegraph Cody0%0%10accepted challenger=heldSince September 2026: 0% → 0%, ±0 points. Inside the 11-point floor: within noise. Read over the models both editions asked.0% → 0%The floor is 11 points of share, measured: how far the models move a leader on their own when the same questions are asked twice with nothing changed. A larger change is movement; a smaller one is noise, and both are shown. Movement is read over the twelve models both editions asked; GPT-6 Luna, Muse Glimmer 30B joined this edition and are in the standing but not yet in the comparison. How the floor is measured
Bars are the share of first choices, 0 to 100Every product with at least 10 labels here. Every product name links to its product page.
Claude Code is made by Anthropic, whose model Claude Haiku 4.5 is in the set. On the four share prompts that model made it the first choice zero times of 4; the other thirteen models zero times of 52. Lab treatment is defined on the method page; the row is marked, not excluded.
All ten head-to-head pages: the top five products, each against each
One product takes 75% of first choices here, so the chart would put eight markers in one corner and one at the far edge. The two measurements it plots are columns in the standing above: share, and the negative label rate. Zero products carry a negative rate above 25% in this category.
What they warned about
Thirteen of fourteen models held their first choice under the paraphrase. Qwen 3.7 Flash changed. A high negative share on a product with few labels is a warning. A low share on a product with many labels is salience, not sentiment.
Cursor19%
11 of 57 labels negative · 8 of 14 models
“Cursor says it sends prompts and code context to model providers like OpenAI, Anthropic, and Google... If you're on an individual plan, the DPA does not apply.” GPT-5.4 mini, negative prompt
GitHub Copilot11%
8 of 73 labels negative · 7 of 14 models
“GitHub says it may use Copilot interaction data, including prompts, suggestions, and code snippets, to train and improve models for individual subscribers.” GPT-5.4 mini, negative prompt
Amazon Q Developer16%
6 of 38 labels negative · 5 of 14 models · 1 hard negative
“Lower Priority/Avoid if Possible: Amazon Q Developer ... End-of-support looming ... migrate from Amazon soon” Grok 4.1 Fast, negative prompt
Claude Code14%
4 of 28 labels negative · 4 of 14 models
“**Claude Code when using third-party skill marketplaces.** The ClawHub/ClawHavoc campaign shows high risk from unsanctioned skills.” Muse Glimmer 30B, negative prompt
What they cite
Citations exist only for the models that return a source list: fourteen of the fourteen in this edition, and all six flagship models on the expanded tier.
Sites the answers cite
75 of 84 answers in this category came back with a source list, from 14 of 14 models: citations where the model returns them, or the search results it consulted. 1030 links across 398 sites, every framing counted. Ranked by the number of answers carrying the site or page. 12 of the 252 answers across every segment cited this index's own page for the category; the method page measures whether that reading tilts an answer.
25 answers · 34 citations · 12 models
vendor site · Augment Code18 answers · 31 citations · 8 models
18 answers · 23 citations · 10 models
17 answers · 18 citations · 8 models
16 answers · 18 citations · 10 models
15 answers · 19 citations · 8 models
vendor site · GitHub14 answers · 23 citations · 7 models
vendor site · Rework14 answers · 16 citations · 8 models
vendor site · GitHub13 answers · 16 citations · 6 models
13 answers · 13 citations · 8 models
12 answers · 14 citations · 8 models
vendor site · G212 answers · 14 citations · 8 models
Pages the answers cite
The ten pages named in the most answers, by full address. A page here is one the models returned with a recommendation, not one the index endorses.
14 answers · 9 models
14 answers · 7 models
12 answers · 7 models
12 answers · 8 models
12 answers · 8 models
11 answers · 8 models
10 answers · 8 models
10 answers · 6 models
9 answers · 6 models
9 answers · 7 models
Search against answers
Each company's standing in the answers beside its site's footprint in Google search, one row a site: the products the models named on it with their shares, and the share they add up to; monthly searches on Google, and DataForSEO's estimate of AI search demand (modeled from search signals, directional, not a count of queries to any assistant), for the most-searched of the company's and its products' names (the name is in each row's hover text); estimated monthly organic visits to the site; and its best position in Google's top ten for “ai coding assistant”, “ai coding assistants”, “best ai coding assistant”. US estimates from DataForSEO and Google's Ads Transparency Center. A small company's site, or a mid-sized company's site for its flagship, is marked company; a product on a large parent's site (Google, Microsoft) has no site figures. A column with no figures for this category is left out, and an empty cell means none were seen, not none exist. Two measurements side by side: neither is read as the cause of the other.
| Company and products | Share | Own site cited | Name searches, Google | AI search demand, est. | Organic visits | Paid search a month, est. | Google ads, last 30 days | Best Google position |
|---|---|---|---|---|---|---|---|---|
| GitHub github.com GitHub Copilot 75% | 75% | 216 | 1,500,000 | 31,479 | 42,019,137 company | $81 | 14 | |
| Windsurf windsurf.com | 14% | 91 | 60,500 | 4,501 | 21,603 | 0 | ||
| Cursor cursor.com | 4% | 167 | 301,000 | 22,500 | 350,718 | $28 | 119 | 6 |
| Amazon Q amazon.com Amazon Q Developer 4% | 4% | 98 | 9,900 | 738 | 1,681,997,879 company | $106,125 | 115 | |
| Tabnine tabnine.com | 2% | 108 | 2,900 | 62 | 15,577 | 0 | ||
| Kilo Code kilo.ai | 2% | 8 | ||||||
| Claude Code claude.com | 0% | 81 | 550,000 | 9,804 | 3,090,945 | $1,889 | 120 | |
| Sourcegraph Cody sourcegraph.com | 0% | 28 | 880 | 14 | 10,627 | 14 | ||
| JetBrains AI Assistant jetbrains.com | 0% | 18 | 880 | 24 | 485,270 | 120 | ||
| Gemini Code Assist | 0% | 6,600 | 167 |
Google's top ten for the three searches hold 24 sites; 1 of them are among the sites the models cited here (augmentcode.com). The first result for “ai coding assistant” is augmentcode.com.
How they answered
Six framings of the same buying question, each sent to every model in a fresh session with search on. One row per model, so a row shows whether it held its answer under rewording, what it named when cost was the constraint, and what it argued against. Computed from the raw judge labels.ShowHide
How they answered
Six framings of the same buying question, each sent to every model in a fresh session with search on. One row per model, so a row shows whether it held its answer under rewording, what it named when cost was the constraint, and what it argued against. Computed from the raw judge labels.
ShowHideModelDirect“What is the best AI coding assistant for a mid-market B2B company?”Paraphrase“Which AI pair programming tool would you recommend to a mid-sized B2B company?”Comparative“What are the top AI coding assistants and how do they differ?”Budget-constrained“What is the best AI coding assistant for a company with a limited budget?”Scale-constrained“We are a 500 person company evaluating an AI coding assistant. What should we look at?”Negative“Which AI coding assistants should I avoid or be cautious about?”
Claude Haiku 4.5GitHub CopilotGitHub CopilotHeldCursorGitHub Copilot, Windsurfagainst: ChatGPTno first choiceagainst: Bolt.new, CodeGenie, Cursor, GitHub Copilot
One alternative
JetBrains AI AssistantThree alternatives
Amazon Q Developer, JetBrains AI Assistant, TabnineFive alternatives
Augment Code, Claude Code, GitHub Copilot, Kiro, TabnineTwo alternatives
Continue, LocalAIGPT-5.4 miniGitHub CopilotGitHub CopilotHeldGitHub CopilotGitHub Copilotagainst: Cursorno first choiceagainst: Claude, Cursor, GitHub Copilot
Two alternatives
Cursor, JetBrains AI AssistantTwo alternatives
JetBrains AI Assistant, Sourcegraph CodyFour alternatives
Claude Code, Cursor, Gemini Code Assist, OpenAI CodexTwo alternatives
Amazon Q Developer, JetBrains AI AssistantGemini 3.5 FlashGitHub CopilotGitHub CopilotHeldCursorKilo Codeagainst: ContinueCursor, GitHub Copilotagainst: Aider, Amazon Q Developer, Augment Code, Claude Code, Cursor, Windsurf
Three alternatives
Augment Code, Cursor, DevinThree alternatives
Cursor, Tabnine, WindsurfSix alternatives
Aider, Claude Code, Continue, GitHub Copilot, Tabnine, WindsurfFour alternatives
Cline, Devin, GitHub Copilot, TabbyTwo alternatives
Claude Code, TabninePerplexity SonarGitHub CopilotGitHub CopilotHeldno first choiceGitHub Copilotno first choicenothing named
Four alternatives
Augment Code, Claude Code, Cursor, Kilo CodeTwo alternatives
Cursor, TabnineSeven alternatives
Aider, Amazon Q Developer, Claude Code, Cursor, Gemini Code Assist, GitHub Copilot, Replit AgentThree alternatives
Cursor, Kilo Code, WindsurfGrok 4.1 FastGitHub CopilotGitHub CopilotHeldGitHub CopilotWindsurfagainst: Cursor, GitHub Copilotno first choiceagainst: Amazon Q Developer, Cursor, GitHub Copilot, Replit AI
Four alternatives
Amazon Q Developer, Cursor, Gemini Code Assist, TabnineTwo alternatives
Cursor, WindsurfThree alternatives
Claude Code, Cursor, WindsurfTwo alternatives
Amazon Q Developer, TabnineThree alternatives
GitHub Copilot, Sourcegraph Cody, TabnineMistral SmallGitHub CopilotGitHub CopilotHeldGitHub CopilotGitHub Copilotno first choiceagainst: Aider, Claude Code, Copilot Workspace, Cursor, Devin, Zed
Two alternatives
Amazon Q Developer, JetBrains AI AssistantFour alternatives
Amazon Q Developer, Replit AI, Sourcegraph Cody, TabnineThree alternatives
Amazon Q Developer, Windsurf, opencodeDeepSeek V4 FlashGitHub CopilotGitHub CopilotHeldClaude Code, CursorGitHub CopilotGitHub Copilotagainst: Q Developeragainst: ChatGPT, GitHub Copilot, Google Gemini, Kilo Code, Trae
Two alternatives
Cursor, WindsurfThree alternatives
Cursor, Tabnine, WindsurfSix alternatives
Amazon Q Developer, Continue, GitHub Copilot, OpenAI Codex, Tabnine, WindsurfTwo alternatives
OpenAI Codex, WindsurfOne alternative
Claude CodeLlama 4 MaverickGitHub CopilotGitHub CopilotHeldno first choiceWindsurfno first choicenothing named
One alternative
Claude CodeTwo alternatives
Cursor, JetBrains AI AssistantFour alternatives
Cursor, GitHub Copilot, Kimi Code, opencodeQwen 3.7 FlashCursor, GitHub CopilotGitHub CopilotChangedno first choiceWindsurfno first choicenothing named
One alternative
Claude CodeTwo alternatives
Amazon Q Developer, CursorFour alternatives
Claude Code, Cursor, GitHub Copilot, WindsurfThree alternatives
Amazon Q Developer, Cursor, GitHub CopilotKimi K2GitHub CopilotGitHub CopilotHeldGitHub CopilotGitHub Copilot, Windsurfno first choiceagainst: Amazon Q Developer, Cursor, GitHub Copilot
Three alternatives
Amazon Q Developer, Cursor, TabnineOne alternative
Amazon Q DeveloperFive alternatives
ChatGPT, Claude Code, Cursor, Tabnine, WindsurfThree alternatives
Amazon Q Developer, Continue, CursorGLM 4.7 FlashXGitHub CopilotGitHub CopilotHeldGitHub Copilotagainst: Microsoft CopilotWindsurfno first choicenothing named
Three alternatives
Cursor, Tabnine, WindsurfOne alternative
TabnineFive alternatives
Amazon CodeWhisperer → Amazon Q Developer, Claude Code, Cursor, Sourcegraph Cody, TabnineThree alternatives
Amazon Q Developer, GitHub Copilot, TabnineMiniMax M2.5GitHub CopilotGitHub CopilotHeldGitHub CopilotAmazon Q Developer, TabnineGitHub Copilotnothing named
Two alternatives
Cursor, WindsurfThree alternatives
Amazon Q Developer, Cursor, TabnineTwo alternatives
Claude Code, CursorOne alternative
CursorOne alternative
Claude CodeGPT-6 LunaGitHub CopilotGitHub CopilotHeldGitHub CopilotGitHub Copilotagainst: Cursorno first choiceagainst: Amazon Q Developer, Augment Code, Claude Code, Cursor, GitHub Copilot, Google Antigravity, Windsurf
Two alternatives
Cursor, Gemini Code AssistOne alternative
CursorSix alternatives
Claude Code, Cursor, Devin, Google Antigravity, OpenAI Codex, WindsurfOne alternative
Gemini Code AssistMuse Glimmer 30BGitHub CopilotGitHub CopilotHeldClaude Code, CursorAmazon Q Developer, Windsurfno first choiceagainst: Amazon Q Developer, Claude Code, Cursor, GitHub Copilot, OpenAI Codex, OpenAI's GPT models, Replit AI, Windsurf
Three alternatives
Amazon Q Developer, Augment Code, CursorThree alternatives
Amazon Q Developer, Sourcegraph Cody, TabnineThree alternatives
Amazon Q Developer, GitHub Copilot, WindsurfFour alternatives
Cline, Continue, GitHub Copilot, TabnineBold is the first choiceAlternatives are counted; the count opens them.What the answer argued against
The record
One row per call: the version string exactly as returned, whether the model searched, sources cited, and latency. Full answer text is in the free responses file. Download the record
Eighty-four rows: every prompt, every model, every answer.
| Prompt | Model | Version string | Time (UTC) | Searched | Sources | Latency |
|---|---|---|---|---|---|---|
| Direct recommendation | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 07:36 | no | 0 | 5 s |
| Direct recommendation | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 07:42 | yes | 6 | 5 s |
| Direct recommendation | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 13:22 | yes | 28 | 36 s |
| Direct recommendation | Perplexity Sonar | sonar | 2026-10-01 12:46 | yes | 20 | 3 s |
| Direct recommendation | Grok 4.1 Fast | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 11:08 | yes | 24 | 9 s |
| Direct recommendation | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 12:25 | yes | 5 | 2 s |
| Direct recommendation | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 09:19 | yes | 24 | 27 s |
| Direct recommendation | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 13:25 | yes | 5 | 2 s |
| Direct recommendation | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 10:52 | yes | 5 | 19 s |
| Direct recommendation | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 08:52 | yes | 10 | 15 s |
| Direct recommendation | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 10:39 | yes | 24 | 32 s |
| Direct recommendation | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 12:35 | yes | 14 | 18 s |
| Direct recommendation | GPT-6 Luna | gpt-6-luna | 2026-10-01 12:24 | yes | 3 | 13 s |
| Direct recommendation | Muse Glimmer 30B | meta/muse-glimmer-30b via togetherai | 2026-10-01 08:51 | yes | 15 | 36 s |
| Paraphrase | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 13:53 | no | 0 | 5 s |
| Paraphrase | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 12:10 | yes | 5 | 5 s |
| Paraphrase | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 13:45 | yes | 15 | 24 s |
| Paraphrase | Perplexity Sonar | sonar | 2026-10-01 10:30 | yes | 26 | 3 s |
| Paraphrase | Grok 4.1 Fast | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 08:18 | yes | 20 | 5 s |
| Paraphrase | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 09:27 | yes | 5 | 3 s |
| Paraphrase | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 10:05 | yes | 24 | 28 s |
| Paraphrase | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 09:16 | yes | 5 | 1 s |
| Paraphrase | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 08:04 | yes | 10 | 36 s |
| Paraphrase | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 10:36 | yes | 10 | 18 s |
| Paraphrase | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 13:53 | yes | 15 | 109 s |
| Paraphrase | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 12:06 | yes | 10 | 26 s |
| Paraphrase | GPT-6 Luna | gpt-6-luna | 2026-10-01 07:59 | yes | 3 | 18 s |
| Paraphrase | Muse Glimmer 30B | meta/muse-glimmer-30b via togetherai | 2026-10-01 07:38 | yes | 15 | 14 s |
| Comparative | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 09:37 | yes | 9 | 9 s |
| Comparative | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 10:43 | yes | 3 | 7 s |
| Comparative | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 09:17 | yes | 15 | 24 s |
| Comparative | Perplexity Sonar | sonar | 2026-10-01 11:30 | yes | 22 | 5 s |
| Comparative | Grok 4.1 Fast | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 08:38 | yes | 24 | 6 s |
| Comparative | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 08:00 | yes | 10 | 6 s |
| Comparative | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 09:05 | yes | 22 | 36 s |
| Comparative | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 08:23 | yes | 5 | 1 s |
| Comparative | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 13:13 | yes | 15 | 66 s |
| Comparative | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 09:22 | yes | 10 | 26 s |
| Comparative | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 12:15 | yes | 10 | 183 s |
| Comparative | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 07:47 | yes | 10 | 23 s |
| Comparative | GPT-6 Luna | gpt-6-luna | 2026-10-01 09:02 | yes | 6 | 37 s |
| Comparative | Muse Glimmer 30B | meta/muse-glimmer-30b via togetherai | 2026-10-01 08:06 | yes | 25 | 37 s |
| Budget constrained | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 07:53 | no | 0 | 5 s |
| Budget constrained | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 10:06 | yes | 4 | 4 s |
| Budget constrained | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 09:20 | yes | 23 | 29 s |
| Budget constrained | Perplexity Sonar | sonar | 2026-10-01 10:13 | yes | 26 | 3 s |
| Budget constrained | Grok 4.1 Fast | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 13:25 | yes | 19 | 6 s |
| Budget constrained | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 12:23 | yes | 5 | 3 s |
| Budget constrained | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 12:24 | yes | 25 | 24 s |
| Budget constrained | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 11:43 | yes | 5 | 3 s |
| Budget constrained | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 10:46 | no | 0 | 26 s |
| Budget constrained | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 08:42 | yes | 15 | 19 s |
| Budget constrained | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 10:16 | yes | 10 | 138 s |
| Budget constrained | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 11:26 | no | 0 | 7 s |
| Budget constrained | GPT-6 Luna | gpt-6-luna | 2026-10-01 11:09 | yes | 4 | 12 s |
| Budget constrained | Muse Glimmer 30B | meta/muse-glimmer-30b via togetherai | 2026-10-01 08:58 | yes | 13 | 21 s |
| Scale constrained | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 12:10 | no | 0 | 6 s |
| Scale constrained | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 10:14 | no | 0 | 7 s |
| Scale constrained | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 09:36 | yes | 14 | 24 s |
| Scale constrained | Perplexity Sonar | sonar | 2026-10-01 10:52 | yes | 20 | 6 s |
| Scale constrained | Grok 4.1 Fast | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 08:24 | yes | 20 | 8 s |
| Scale constrained | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 08:44 | no | 0 | 10 s |
| Scale constrained | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 12:27 | yes | 22 | 41 s |
| Scale constrained | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 09:47 | yes | 5 | 4 s |
| Scale constrained | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 09:21 | yes | 19 | 48 s |
| Scale constrained | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 12:42 | yes | 15 | 45 s |
| Scale constrained | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 10:19 | yes | 19 | 40 s |
| Scale constrained | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 09:50 | yes | 5 | 16 s |
| Scale constrained | GPT-6 Luna | gpt-6-luna | 2026-10-01 08:34 | yes | 3 | 16 s |
| Scale constrained | Muse Glimmer 30B | meta/muse-glimmer-30b via togetherai | 2026-10-01 11:53 | yes | 13 | 31 s |
| Negative framing | Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 2026-10-01 12:48 | yes | 18 | 10 s |
| Negative framing | GPT-5.4 mini | gpt-5.4-mini-2026-03-17 | 2026-10-01 08:42 | yes | 3 | 4 s |
| Negative framing | Gemini 3.5 Flash | gemini-3.5-flash | 2026-10-01 13:53 | yes | 22 | 20 s |
| Negative framing | Perplexity Sonar | sonar | 2026-10-01 08:28 | yes | 25 | 4 s |
| Negative framing | Grok 4.1 Fast | spacexai/grok-4.1-fast-non-reasoning via vertex | 2026-10-01 09:04 | yes | 25 | 7 s |
| Negative framing | Mistral Small | mistral/mistral-small via mistral | 2026-10-01 09:10 | yes | 5 | 5 s |
| Negative framing | DeepSeek V4 Flash | deepseek/deepseek-v4-flash via deepinfra | 2026-10-01 13:28 | yes | 25 | 28 s |
| Negative framing | Llama 4 Maverick | meta/llama-4-maverick via bedrock | 2026-10-01 07:50 | yes | 5 | 1 s |
| Negative framing | Qwen 3.7 Flash | alibaba/qwen3.7-flash via alibaba | 2026-10-01 08:28 | yes | 10 | 24 s |
| Negative framing | Kimi K2 | moonshotai/kimi-k2 via novita | 2026-10-01 09:15 | yes | 25 | 32 s |
| Negative framing | GLM 4.7 FlashX | zai/glm-4.7-flashx via zai | 2026-10-01 12:19 | no | 0 | 21 s |
| Negative framing | MiniMax M2.5 | minimax/minimax-m2.5 via minimax | 2026-10-01 13:38 | yes | 5 | 18 s |
| Negative framing | GPT-6 Luna | gpt-6-luna | 2026-10-01 09:17 | yes | 4 | 23 s |
| Negative framing | Muse Glimmer 30B | meta/muse-glimmer-30b via togetherai | 2026-10-01 10:56 | yes | 22 | 28 s |
Noise floor in this category
Flips between the edition run and its calibration repeat. Six prompts per model is a small sample; the index-wide floor is the number to trust.
ShowHide
Claude Haiku 4.5
2 of 4 flipped
GPT-5.4 mini
2 of 4 flipped
Gemini 3.5 Flash
4 of 5 flipped
Perplexity Sonar
1 of 4 flipped
Grok 4.1 Fast
3 of 6 flipped
Mistral Small
3 of 5 flipped
DeepSeek V4 Flash
1 of 6 flipped
Llama 4 Maverick
1 of 3 flipped
Qwen 3.7 Flash
3 of 5 flipped
Kimi K2
2 of 6 flipped
GLM 4.7 FlashX
1 of 4 flipped
MiniMax M2.5
3 of 5 flipped
GPT-6 Luna
0 of 4 flipped
Muse Glimmer 30B
3 of 5 flipped
Normalization in this category
Every judgment call made between the raw labels and the numbers above, listed so it is visible and reversible.
ShowHide
Category-scoped readings
Amazon Q read as Amazon Q Developer
Augment read as Augment Code
Copilot read as GitHub Copilot
JetBrains AI read as JetBrains AI Assistant
Replit read as Replit AI
Unresolved, counted raw
Amazon CodeWhisperer → Amazon Q Developer
CodeGenie
CodeLlama
GitHub Copilot Workspace / Business
IBM Bob
Kimi Code
LocalAI
OpenAI's GPT models
PlanToCode
Q Developer
Trae
VS Code + plugins
Discontinued, still offered
No shut-down product was recommended here.