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.
| # | Model | DeepSWE | Agent | Agent arch | Context | Cost | Open | Pairing basis |
|---|---|---|---|---|---|---|---|---|
| 1 | Gemini 4 Argon Google | 77.9 | Gemini CLI | 48 | 1M | not listed | No | vendor harness |
| 2 | GPT-6.1 Sol OpenAI | 75.2 | ChatGPT Codex | 58 | 1.1M | $2/$10 | No | vendor harness |
| 3 | Claude Opus 5.5 Anthropic | 74.2 | Claude Code | 66 | 1M | $4/$20 | No | vendor harness |
| 4 | DeepSeek V4.1 Flash DeepSeek | 74.2 | Aider | 43 | not listed | $0.15/$0.60 | Yes | open weights, open agent |
| 5 | GPT-6 Astra OpenAI | 74.1 | ChatGPT Codex | 58 | 272K+ | $10/$50 | No | vendor harness |
| 6 | Gemini 3.8 Flash Google | 73.7 | Gemini CLI | 48 | 1M | $0.75/$3.75 | No | vendor harness |
| 7 | MiMo-V2.6-Pro Xiaomi | 71.9 | Cursor | 56 | 1M | $0.435/$0.87 | Yes | open weights, IDE agent |
| 8 | Claude Sonnet 5.5 Anthropic | 71.0 | Claude Code | 66 | 1M | $2/$10 | No | vendor harness |
| 9 | Claude Fable 5 Anthropic | 70.0 | Claude Code | 66 | 1M | $10/$50 | No | vendor harness |
| 10 | GLM-5.3 Z.ai | 66.9 | OpenCode | 48 | not listed | not listed | Yes | open weights, open agent |
| 11 | Mistral Large 4 Mistral AI | 61.7 | Aider | 43 | 1M | $0.68/$2.09 | No | open weights, open agent |
| 12 | Qwen3.8-Flash-Next Alibaba | 58.7 | Cline | 47 | not listed | not listed | Yes | open weights, open agent |
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.
| # | Stack | ARscore | Context | Cost | Best fit |
|---|---|---|---|---|---|
| 1 | Claude Code + Opus 4.8 CLI agent | 76.2 | 1M | $5/$25 | deep terminal workflows and complex multi-step refactors |
| 2 | Claude Code + Opus 4.7 CLI agent | 75.3 | 1M | $5/$30 | deep terminal workflows and complex multi-step refactors |
| 3 | ChatGPT Codex + GPT-5.5 CLI agent | 69.4 | 258K | $6/$25 | ChatGPT Codex workflows, terminal automation, and CI-style editing |
| 4 | Claude Code + Sonnet 4.6 CLI agent | 68.5 | 1M | $2/$8 | deep terminal workflows and complex multi-step refactors |
| 5 | ChatGPT Codex + GPT-5.3 Codex CLI agent | 63.2 | 128K | $5/$20 | ChatGPT Codex workflows, terminal automation, and CI-style editing |
| 6 | Cursor + Opus 4.8 IDE agent | 62.9 | 1M | $5/$25 | daily IDE coding with autocomplete and composer workflows |
| 7 | Cursor + GPT-5.5 IDE agent | 58.6 | 258K | $6/$25 | daily IDE coding with autocomplete and composer workflows |
| 8 | Cursor + DeepSeek V4 Pro IDE agent | 56.2 | 1M | $0.44/$0.87 | daily IDE coding with autocomplete and composer workflows |
| 9 | Cursor + Gemini 3.1 Pro IDE agent | 57.2 | 1M+ | $1.5/$6 | daily IDE coding with autocomplete and composer workflows |
| 10 | Cursor + Qwen 3.7 Max IDE agent | 57.1 | 1M | $1.66/$4.95 | daily IDE coding with autocomplete and composer workflows |
| 11 | Cursor + Kimi K2.6 IDE agent | 56.9 | 256K | $0.6/$2.5 | daily IDE coding with autocomplete and composer workflows |
| 12 | Cursor + GPT-5.4 IDE agent | 56.2 | 1M | $5/$20 | daily IDE coding with autocomplete and composer workflows |
| 13 | Cursor + DeepSeek V4 Flash IDE agent | 56.1 | 1M | $0.14/$0.28 | daily IDE coding with autocomplete and composer workflows |
| 14 | Cursor + Gemini 3 Flash IDE agent | 54.1 | 1M+ | $0.5/$3 | daily IDE coding with autocomplete and composer workflows |
| 15 | Devin + GPT-5.5 Cloud agent | 50.4 | 258K | $6/$25 | long-running autonomous cloud tasks |
| 16 | Cline + DeepSeek V4 ProOS Extension agent | 48.3 | 1M | $0.44/$0.87 | MCP-heavy VS Code workflows and bring-your-own-model setups |
| 17 | Cline + Kimi K2.6OS Extension agent | 48.9 | 256K | $0.6/$2.5 | MCP-heavy VS Code workflows and bring-your-own-model setups |
| 18 | Windsurf + GPT-5.5 IDE agent | 48.7 | 258K | $6/$25 | AI-native IDE workflows with Cascade-style multi-file edits |
| 19 | Devin + Sonnet 4.6 Cloud agent | 48.6 | 1M | $2/$8 | long-running autonomous cloud tasks |
| 20 | Aider + DeepSeek V4 ProOS CLI agent | 47.5 | 1M | $0.44/$0.87 | git-native terminal editing and experienced open-source users |
| 21 | Cline + DeepSeek V4 FlashOS Extension agent | 48.2 | 1M | $0.14/$0.28 | MCP-heavy VS Code workflows and bring-your-own-model setups |
| 22 | Aider + Kimi K2.6OS CLI agent | 48.1 | 256K | $0.6/$2.5 | git-native terminal editing and experienced open-source users |
| 23 | Continue + DeepSeek V4 ProOS Extension agent | 46.7 | 1M | $0.44/$0.87 | custom context providers across VS Code and JetBrains |
| 24 | Windsurf + DeepSeek V4 Pro IDE agent | 46.7 | 1M | $0.44/$0.87 | AI-native IDE workflows with Cascade-style multi-file edits |
| 25 | Aider + DeepSeek V4 FlashOS CLI agent | 47.4 | 1M | $0.14/$0.28 | git-native terminal editing and experienced open-source users |
| 26 | Continue + Sonnet 4.6 Extension agent | 47 | 1M | $2/$8 | custom context providers across VS Code and JetBrains |
| 27 | Windsurf + Sonnet 4.6 IDE agent | 47 | 1M | $2/$8 | AI-native IDE workflows with Cascade-style multi-file edits |
| 28 | Copilot + GPT-5.5 IDE agent | 43 | 258K | $6/$25 | GitHub-centered teams and IDE-native suggestions |
| 29 | OpenCode + DeepSeek V4 ProOS CLI agent | 41.9 | 1M | $0.44/$0.87 | provider-flexible terminal coding with open-source control |
| 30 | OpenCode + Kimi K2.6OS CLI agent | 42.5 | 256K | $0.6/$2.5 | provider-flexible terminal coding with open-source control |
| 31 | OpenCode + DeepSeek V4 FlashOS CLI agent | 41.9 | 1M | $0.14/$0.28 | provider-flexible terminal coding with open-source control |
| 32 | Copilot + GPT-5.4 IDE agent | 41.2 | 1M | $5/$20 | GitHub-centered teams and IDE-native suggestions |
| 33 | Claude Code + Opus Mythos unknown 50% hype haircut: no reliable public community test data yet. | 40.4 | 1M | TBD | deep terminal workflows and complex multi-step refactors |
| 34 | Continue + Mistral Large 3OS Extension agent | 40.1 | 128K | $2/$6 | custom context providers across VS Code and JetBrains |
| 35 | Cline + Qwen 3.6OS Extension agent | 39.7 | 128K | $0.3/$0.6 | MCP-heavy VS Code workflows and bring-your-own-model setups |
| 36 | Aider + Qwen 3.6OS CLI agent | 39 | 128K | $0.3/$0.6 | git-native terminal editing and experienced open-source users |
| 37 | Oh-My-Codex + GPT-5.5 CLI agent | 38.8 | 258K | $6/$25 | open-source Codex enhancements |
| 38 | Cline + Llama 4 MaverickOS Extension agent | 37.8 | 1M | $0.8/$2.2 | MCP-heavy VS Code workflows and bring-your-own-model setups |
| 39 | Oh-My-Codex + Sonnet 4.6 CLI agent | 37.4 | 1M | $2/$8 | open-source Codex enhancements |
| 40 | Aider + Llama 4 MaverickOS CLI agent | 37.2 | 1M | $0.8/$2.2 | git-native terminal editing and experienced open-source users |
| 41 | Qwen Code + Qwen 3.7 Max CLI agent | 36.2 | 1M | $1.66/$4.95 | Qwen-first open-source CLI workflows |
| 42 | Goose + DeepSeek V4 ProOS CLI agent | 34.7 | 1M | $0.44/$0.87 | MCP-compatible local automation |
| 43 | OpenCode + Qwen 3.6OS CLI agent | 34.5 | 128K | $0.3/$0.6 | provider-flexible terminal coding with open-source control |
| 44 | Goose + DeepSeek V4 FlashOS CLI agent | 34 | 1M | $0.14/$0.28 | MCP-compatible local automation |
| 45 | Gemini CLI + Gemini 3.1 Pro CLI agent | 32.2 | 1M+ | $1.5/$6 | legacy Google CLI workflows |
| 46 | Hermes Agent + DeepSeek V4 ProOS CLI agent | 31.4 | 1M | $0.44/$0.87 | skills-based terminal experimentation |
| 47 | Hermes Agent + DeepSeek V4 FlashOS CLI agent | 30.8 | 1M | $0.14/$0.28 | skills-based terminal experimentation |
| 48 | Gemini CLI + Gemini 3 Flash CLI agent | 30.5 | 1M+ | $0.5/$3 | legacy Google CLI workflows |
| 49 | Qwen Code + Qwen 3.6OS CLI agent | 29.3 | 128K | $0.3/$0.6 | Qwen-first open-source CLI workflows |
| 50 | Antigravity CLI + Gemini 3.1 Pro CLI agent | 28.2 | 1M+ | $1.5/$6 | Google terminal-agent exploration |
| 51 | Pi Agent + DeepSeek V4 FlashOS CLI agent | 27.7 | 1M | $0.14/$0.28 | minimal-token terminal use |
| 52 | Replit Agent + DeepSeek V4 Pro Browser agent | 27.4 | 1M | $0.44/$0.87 | browser-based app building |
| 53 | Replit Agent + DeepSeek V4 Flash Browser agent | 26.9 | 1M | $0.14/$0.28 | browser-based app building |
| 54 | Antigravity CLI + Gemini 3 Flash CLI agent | 26.7 | 1M+ | $0.5/$3 | Google terminal-agent exploration |
| 55 | Goose + Llama 4 MaverickOS CLI agent | 26.7 | 1M | $0.8/$2.2 | MCP-compatible local automation |
| 56 | Amazon Q Developer + DeepSeek V3.2 IDE agent | 26.6 | 128K | $0.25/$0.38 | AWS-heavy enterprise teams |
| 57 | Pi Agent + DeepSeek V3.2OS CLI agent | 24.5 | 128K | $0.25/$0.38 | minimal-token terminal use |
| 58 | Hermes Agent + Llama 4 MaverickOS CLI agent | 24.2 | 1M | $0.8/$2.2 | skills-based terminal experimentation |
| 59 | Pi Agent + Qwen 3.6OS CLI agent | 22.8 | 128K | $0.3/$0.6 | minimal-token terminal use |
| 60 | Replit Agent + Llama 4 Maverick Browser agent | 21.1 | 1M | $0.8/$2.2 | browser-based app building |
| 61 | Bolt.new + GPT-5.5 Browser agent | 20.7 | 258K | $6/$25 | browser-based full-stack prototypes |
| 62 | Bolt.new + Sonnet 4.6 Browser agent | 19.9 | 1M | $2/$8 | browser-based full-stack prototypes |
| 63 | Tabnine + Llama 4 Maverick IDE agent | 17.4 | 1M | $0.8/$2.2 | enterprise code completion and private deployments |
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.
Best Open Source AI Coding Agents
Open-source agents are strongest when you want provider choice, local-model support, and transparent agent behavior.
Best AI Coding Agents for VS Code
IDE-focused agents are best for daily coding, codebase navigation, completions, and fast multi-file edits.
Best Browser AI Coding Agents
Browser coding agents are useful when you want to build, preview, and ship without managing a local development environment.