~ / best / highest-score-ai-coding-stacks

Highest Score AI Coding Stacks

Two tables: the current leading models each paired with an agent that runs them, then the June 2026 stack snapshot. 12 current pairs, 63 archived stacks.

Current models, paired with an agent that runs them

Ordered by the model's DeepSWE 1.1 score. There is no stack-level number here, because none of these combinations has been measured end-to-end — see the note below.

#ModelDeepSWEAgentAgent archContextCostOpenPairing basis
1Gemini 4 Argon
Google
77.9Gemini CLI481Mnot listedNovendor harness
2GPT-6.1 Sol
OpenAI
75.2ChatGPT Codex581.1M$2/$10Novendor harness
3Claude Opus 5.5
Anthropic
74.2Claude Code661M$4/$20Novendor harness
4DeepSeek V4.1 Flash
DeepSeek
74.2Aider43not listed$0.15/$0.60Yesopen weights, open agent
5GPT-6 Astra
OpenAI
74.1ChatGPT Codex58272K+$10/$50Novendor harness
6Gemini 3.8 Flash
Google
73.7Gemini CLI481M$0.75/$3.75Novendor harness
7MiMo-V2.6-Pro
Xiaomi
71.9Cursor561M$0.435/$0.87Yesopen weights, IDE agent
8Claude Sonnet 5.5
Anthropic
71.0Claude Code661M$2/$10Novendor harness
9Claude Fable 5
Anthropic
70.0Claude Code661M$10/$50Novendor harness
10GLM-5.3
Z.ai
66.9OpenCode48not listednot listedYesopen weights, open agent
11Mistral Large 4
Mistral AI
61.7Aider431M$0.68/$2.09Noopen weights, open agent
12Qwen3.8-Flash-Next
Alibaba
58.7Cline47not listednot listedYesopen weights, open agent
Why there is no ARscore column above. ARscore was a stack-level figure measured in June 2026 against a model set that no longer includes a single one of the current leaders. Rather than extrapolate it onto models it was never measured on, this table shows the two layers separately and orders by the measured one. The June stacks follow, unchanged, as a snapshot.

June 2026 stack snapshot (ARscore, archived)

Measured 21 June 2026. Kept for the record; the model layer has moved a long way since, and Opus 4.8 — the top row here — now sits 31st on the DeepSWE table.

#StackARscoreContextCostBest fit
1Claude Code + Opus 4.8
CLI agent
76.21M$5/$25deep terminal workflows and complex multi-step refactors
2Claude Code + Opus 4.7
CLI agent
75.31M$5/$30deep terminal workflows and complex multi-step refactors
3ChatGPT Codex + GPT-5.5
CLI agent
69.4258K$6/$25ChatGPT Codex workflows, terminal automation, and CI-style editing
4Claude Code + Sonnet 4.6
CLI agent
68.51M$2/$8deep terminal workflows and complex multi-step refactors
5ChatGPT Codex + GPT-5.3 Codex
CLI agent
63.2128K$5/$20ChatGPT Codex workflows, terminal automation, and CI-style editing
6Cursor + Opus 4.8
IDE agent
62.91M$5/$25daily IDE coding with autocomplete and composer workflows
7Cursor + GPT-5.5
IDE agent
58.6258K$6/$25daily IDE coding with autocomplete and composer workflows
8Cursor + DeepSeek V4 Pro
IDE agent
56.21M$0.44/$0.87daily IDE coding with autocomplete and composer workflows
9Cursor + Gemini 3.1 Pro
IDE agent
57.21M+$1.5/$6daily IDE coding with autocomplete and composer workflows
10Cursor + Qwen 3.7 Max
IDE agent
57.11M$1.66/$4.95daily IDE coding with autocomplete and composer workflows
11Cursor + Kimi K2.6
IDE agent
56.9256K$0.6/$2.5daily IDE coding with autocomplete and composer workflows
12Cursor + GPT-5.4
IDE agent
56.21M$5/$20daily IDE coding with autocomplete and composer workflows
13Cursor + DeepSeek V4 Flash
IDE agent
56.11M$0.14/$0.28daily IDE coding with autocomplete and composer workflows
14Cursor + Gemini 3 Flash
IDE agent
54.11M+$0.5/$3daily IDE coding with autocomplete and composer workflows
15Devin + GPT-5.5
Cloud agent
50.4258K$6/$25long-running autonomous cloud tasks
16Cline + DeepSeek V4 ProOS
Extension agent
48.31M$0.44/$0.87MCP-heavy VS Code workflows and bring-your-own-model setups
17Cline + Kimi K2.6OS
Extension agent
48.9256K$0.6/$2.5MCP-heavy VS Code workflows and bring-your-own-model setups
18Windsurf + GPT-5.5
IDE agent
48.7258K$6/$25AI-native IDE workflows with Cascade-style multi-file edits
19Devin + Sonnet 4.6
Cloud agent
48.61M$2/$8long-running autonomous cloud tasks
20Aider + DeepSeek V4 ProOS
CLI agent
47.51M$0.44/$0.87git-native terminal editing and experienced open-source users
21Cline + DeepSeek V4 FlashOS
Extension agent
48.21M$0.14/$0.28MCP-heavy VS Code workflows and bring-your-own-model setups
22Aider + Kimi K2.6OS
CLI agent
48.1256K$0.6/$2.5git-native terminal editing and experienced open-source users
23Continue + DeepSeek V4 ProOS
Extension agent
46.71M$0.44/$0.87custom context providers across VS Code and JetBrains
24Windsurf + DeepSeek V4 Pro
IDE agent
46.71M$0.44/$0.87AI-native IDE workflows with Cascade-style multi-file edits
25Aider + DeepSeek V4 FlashOS
CLI agent
47.41M$0.14/$0.28git-native terminal editing and experienced open-source users
26Continue + Sonnet 4.6
Extension agent
471M$2/$8custom context providers across VS Code and JetBrains
27Windsurf + Sonnet 4.6
IDE agent
471M$2/$8AI-native IDE workflows with Cascade-style multi-file edits
28Copilot + GPT-5.5
IDE agent
43258K$6/$25GitHub-centered teams and IDE-native suggestions
29OpenCode + DeepSeek V4 ProOS
CLI agent
41.91M$0.44/$0.87provider-flexible terminal coding with open-source control
30OpenCode + Kimi K2.6OS
CLI agent
42.5256K$0.6/$2.5provider-flexible terminal coding with open-source control
31OpenCode + DeepSeek V4 FlashOS
CLI agent
41.91M$0.14/$0.28provider-flexible terminal coding with open-source control
32Copilot + GPT-5.4
IDE agent
41.21M$5/$20GitHub-centered teams and IDE-native suggestions
33Claude Code + Opus Mythos unknown
50% hype haircut: no reliable public community test data yet.
40.41MTBDdeep terminal workflows and complex multi-step refactors
34Continue + Mistral Large 3OS
Extension agent
40.1128K$2/$6custom context providers across VS Code and JetBrains
35Cline + Qwen 3.6OS
Extension agent
39.7128K$0.3/$0.6MCP-heavy VS Code workflows and bring-your-own-model setups
36Aider + Qwen 3.6OS
CLI agent
39128K$0.3/$0.6git-native terminal editing and experienced open-source users
37Oh-My-Codex + GPT-5.5
CLI agent
38.8258K$6/$25open-source Codex enhancements
38Cline + Llama 4 MaverickOS
Extension agent
37.81M$0.8/$2.2MCP-heavy VS Code workflows and bring-your-own-model setups
39Oh-My-Codex + Sonnet 4.6
CLI agent
37.41M$2/$8open-source Codex enhancements
40Aider + Llama 4 MaverickOS
CLI agent
37.21M$0.8/$2.2git-native terminal editing and experienced open-source users
41Qwen Code + Qwen 3.7 Max
CLI agent
36.21M$1.66/$4.95Qwen-first open-source CLI workflows
42Goose + DeepSeek V4 ProOS
CLI agent
34.71M$0.44/$0.87MCP-compatible local automation
43OpenCode + Qwen 3.6OS
CLI agent
34.5128K$0.3/$0.6provider-flexible terminal coding with open-source control
44Goose + DeepSeek V4 FlashOS
CLI agent
341M$0.14/$0.28MCP-compatible local automation
45Gemini CLI + Gemini 3.1 Pro
CLI agent
32.21M+$1.5/$6legacy Google CLI workflows
46Hermes Agent + DeepSeek V4 ProOS
CLI agent
31.41M$0.44/$0.87skills-based terminal experimentation
47Hermes Agent + DeepSeek V4 FlashOS
CLI agent
30.81M$0.14/$0.28skills-based terminal experimentation
48Gemini CLI + Gemini 3 Flash
CLI agent
30.51M+$0.5/$3legacy Google CLI workflows
49Qwen Code + Qwen 3.6OS
CLI agent
29.3128K$0.3/$0.6Qwen-first open-source CLI workflows
50Antigravity CLI + Gemini 3.1 Pro
CLI agent
28.21M+$1.5/$6Google terminal-agent exploration
51Pi Agent + DeepSeek V4 FlashOS
CLI agent
27.71M$0.14/$0.28minimal-token terminal use
52Replit Agent + DeepSeek V4 Pro
Browser agent
27.41M$0.44/$0.87browser-based app building
53Replit Agent + DeepSeek V4 Flash
Browser agent
26.91M$0.14/$0.28browser-based app building
54Antigravity CLI + Gemini 3 Flash
CLI agent
26.71M+$0.5/$3Google terminal-agent exploration
55Goose + Llama 4 MaverickOS
CLI agent
26.71M$0.8/$2.2MCP-compatible local automation
56Amazon Q Developer + DeepSeek V3.2
IDE agent
26.6128K$0.25/$0.38AWS-heavy enterprise teams
57Pi Agent + DeepSeek V3.2OS
CLI agent
24.5128K$0.25/$0.38minimal-token terminal use
58Hermes Agent + Llama 4 MaverickOS
CLI agent
24.21M$0.8/$2.2skills-based terminal experimentation
59Pi Agent + Qwen 3.6OS
CLI agent
22.8128K$0.3/$0.6minimal-token terminal use
60Replit Agent + Llama 4 Maverick
Browser agent
21.11M$0.8/$2.2browser-based app building
61Bolt.new + GPT-5.5
Browser agent
20.7258K$6/$25browser-based full-stack prototypes
62Bolt.new + Sonnet 4.6
Browser agent
19.91M$2/$8browser-based full-stack prototypes
63Tabnine + Llama 4 Maverick
IDE agent
17.41M$0.8/$2.2enterprise code completion and private deployments
How to read this: Agent pages score the wrapper and workflow. Stack pages score agent plus model combinations.

Related rankings

Other useful paths through the dataset.

Best CLI AI Coding Agents

Terminal-first coding agents are best when you want repo-level edits, scriptable workflows, and less IDE lock-in.

11 entries

Best Open Source AI Coding Agents

Open-source agents are strongest when you want provider choice, local-model support, and transparent agent behavior.

9 entries

Best AI Coding Agents for VS Code

IDE-focused agents are best for daily coding, codebase navigation, completions, and fast multi-file edits.

7 entries

Best Browser AI Coding Agents

Browser coding agents are useful when you want to build, preview, and ship without managing a local development environment.

3 entries