Benchmark leaderboard - 25 entries
Aider Polyglot leaderboard
Aider's polyglot benchmark tests LLMs on 225 challenging Exercism coding exercises across C++, Go, Java, JavaScript, Python, and Rust. As of October 17, 2026, gpt-5 (high) ranks first on the Aider Polyglot board with 88.0%.
Last verified from aider.chat. Read the caveats
Aider Polyglot
25 entries, in the board's order
| 1 | gpt-5 (high) Aider, edit format diff, command `aider --model openai/gpt-5`, reasoning effort high, percent correct (pass rate 2) on 225 cases, correct edit format 91.6%, cost $29.08 Official leaderboard | 88.0% |
| 2 | gpt-5 (medium) Aider, edit format diff, command `aider --model openai/gpt-5`, reasoning effort medium, correct edit format 88.4%, cost $17.69 Official leaderboard | 86.7% |
| 3 | o3-pro (high) Aider, edit format diff, command `aider --model o3-pro`, correct edit format 97.8%, cost $146.32 Official leaderboard | 84.9% |
| 4 | gemini-2.5-pro-preview-06-05 (32k think) Aider, edit format diff-fenced, command `aider --model gemini/gemini-2.5-pro-preview-06-05 --thinking-tokens 32k`, correct edit format 99.6%, cost $49.88 Official leaderboard | 83.1% |
| 5 | gpt-5 (low) Aider, edit format diff, command `aider --model openai/gpt-5`, reasoning effort low, correct edit format 86.7%, cost $10.37 Official leaderboard | 81.3% |
| 6 | o3 (high) Aider, edit format diff, command `aider --model o3 --reasoning-effort high`, correct edit format 94.7%, cost $21.23 Official leaderboard | 81.3% |
| 7 | grok-4 (high) Aider, edit format diff, command `aider --model openrouter/x-ai/grok-4`, correct edit format 97.3%, cost $59.62 Official leaderboard | 79.6% |
| 8 | gemini-2.5-pro-preview-06-05 (default think) Aider, edit format diff-fenced, command `aider --model gemini/gemini-2.5-pro-preview-06-05`, correct edit format 100.0%, cost $45.6 Official leaderboard | 79.1% |
| 9 | o3 (high) + gpt-4.1 Aider, edit format architect (o3 high as architect, gpt-4.1 as editor), command `aider --model o3`, correct edit format 100.0%, cost $17.55 Official leaderboard | 78.2% |
| 10 | o3 Aider, edit format diff, command `aider --model o3`, correct edit format 93.8%, cost $13.75 Official leaderboard | 76.9% |
| 11 | Gemini 2.5 Pro Preview 05-06 Aider, edit format diff-fenced, command `aider --model gemini/gemini-2.5-pro-preview-05-06`, correct edit format 97.3%, cost $37.41 Official leaderboard | 76.9% |
| 12 | DeepSeek-V3.2-Exp (Reasoner) Aider, edit format diff, command `aider --model deepseek/deepseek-reasoner`, correct edit format 97.3%, cost $1.3 Official leaderboard | 74.2% |
| 13 | Gemini 2.5 Pro Preview 03-25 Aider, edit format diff-fenced, command `aider --model gemini/gemini-2.5-pro-preview-03-25`, correct edit format 92.4%, cost not shown Official leaderboard | 72.9% |
| 14 | claude-opus-4-20250514 (32k thinking) Aider, edit format diff, command `aider --model claude-opus-4-20250514`, correct edit format 97.3%, cost $65.75 Official leaderboard | 72.0% |
| 15 | o4-mini (high) Aider, edit format diff, command `aider --model o4-mini`, correct edit format 90.7%, cost $19.64 Official leaderboard | 72.0% |
| 16 | DeepSeek R1 (0528) Aider, edit format diff, command `aider --model deepseek/deepseek-reasoner`, correct edit format 94.6%, cost $4.8 Official leaderboard | 71.4% |
| 17 | claude-opus-4-20250514 (no think) Aider, edit format diff, command `aider --model claude-opus-4-20250514`, correct edit format 98.7%, cost $68.63 Official leaderboard | 70.7% |
| 18 | DeepSeek-V3.2-Exp (Chat) Aider, edit format diff, command `aider --model deepseek/deepseek-chat`, correct edit format 98.2%, cost $0.88 Official leaderboard | 70.2% |
| 19 | claude-3-7-sonnet-20250219 (32k thinking tokens) Aider, edit format diff, command `aider --model anthropic/claude-3-7-sonnet-20250219 --thinking-tokens 32k`, correct edit format 97.8%, cost $36.83 Official leaderboard | 64.9% |
| 20 | DeepSeek R1 + claude-3-5-sonnet-20241022 Aider, edit format architect, command `aider --architect --model r1 --editor-model sonnet`, correct edit format 100.0%, cost $13.29 Official leaderboard | 64.0% |
| 21 | o1-2024-12-17 (high) Aider, edit format diff, command `aider --model openrouter/openai/o1`, correct edit format 91.5%, cost $186.5 (224 test cases recorded) Official leaderboard | 61.7% |
| 22 | claude-sonnet-4-20250514 (32k thinking) Aider, edit format diff, command `aider --model claude-sonnet-4-20250514`, correct edit format 97.3%, cost $26.58 Official leaderboard | 61.3% |
| 23 | claude-3-7-sonnet-20250219 (no thinking) Aider, edit format diff, command `aider --model sonnet`, correct edit format 93.3%, cost $17.72 Official leaderboard | 60.4% |
| 24 | o3-mini (high) Aider, edit format diff, command `aider --model o3-mini --reasoning-effort high`, correct edit format 93.3%, cost $18.16 Official leaderboard | 60.4% |
| 25 | Qwen3 235B A22B diff, no think, Alibaba API Aider, edit format diff, command `aider --model openai/qwen3-235b-a22b`, correct edit format 92.9%, cost not shown Official leaderboard | 59.6% |
About this board
- The official Aider polyglot leaderboard (read 2026-10-17) says 'last updated November 20, 2025'; its newest entry is dated 2025-10-03. Newer frontier models (for example, anything released after October 2025) have no entry on the official board, so the top 25 reflects mid-2025 models only.
- Raw data cross-checked against https://github.com/Aider-AI/aider/blob/main/aider/website/_data/polyglot_leaderboard.yml (69 entries); scores match the rendered page.
- Ties are kept in the page's displayed order: gpt-5 (low) / o3 (high) at 81.3%, o3 / Gemini 2.5 Pro Preview 05-06 at 76.9%, claude-opus-4 32k / o4-mini (high) at 72.0%, claude-3-7-sonnet no thinking / o3-mini (high) at 60.4%. Rank 26 is Kimi K2 at 59.1%.
FAQ
What does Aider Polyglot measure?
Aider's polyglot benchmark tests LLMs on 225 challenging Exercism coding exercises across C++, Go, Java, JavaScript, Python, and Rust.
Which model leads Aider Polyglot?
As of October 17, 2026, gpt-5 (high) ranks first on the Aider Polyglot board with 88.0%. That is the board's own order, not a Developers Digest ranking.
Where do these Aider Polyglot scores come from?
All 25 entries come from the official leaderboard at aider.chat. Every entry links to its source, which was read on October 17, 2026, and lists the setting it was run under.
Related reading
Sources
Last verified . Each row links to the page its score was copied from.
Newsletter
Get new benchmark results as they land
One email per week. New model releases, the leaderboard moves that matter and what they mean for what you build.