{ }AICoder
English 中文
June 2026

AI Coder Bench

Real-world coding benchmark rankings. 50 tasks across bug fixing, feature building, refactoring, system design, and debug & explain.

6Boards
49Models ranked
777Tasks
July 2026Period
2026-07-18Last update
●We fetch raw rankings from 6 public benchmark providers directly — no re-weighting, no composite score. Click "Open official board" on each card to go to the source.

📌 4 lenses · 6 official boards

Each card previews a benchmark provider's Top 5. Click to expand full rankings, or jump straight to the official page.

LMArena ★
Hundreds of thousands of human blind votes (Elo)
Chat
Metric:Arena Elo· 12 models
  1. #1Claude Fable 51631
  2. #2Claude Opus 4.81581
  3. #3Grok 4.51558
  4. #4Claude Sonnet 4.61544
  5. #5Kimi K2.61519
Official ↗
Artificial Analysis
Composite Intelligence Index over 20+ benchmarks
Overall
Metric:AA Index· 46 models
  1. #1Claude Fable 559.9
  2. #2GPT-5.6 Sol58.9
  3. #3Kimi K357.1
  4. #4Claude Opus 4.855.7
  5. #5GPT-5.6 Terra55
Official ↗
LiveBench
Contamination-resistant benchmark, refreshed monthly
Reasoning
Metric:Avg score%· 23 models
  1. #1GPT-5.6 Sol82.9%
  2. #2Claude Fable 581.4%
  3. #3GPT-5.580.5%
  4. #4GPT-5.6 Terra80.3%
  5. #5Claude Opus 4.879.8%
Official ↗
Aider Polyglot
133 real coding tasks across 6 languages
Coding
Metric:Pass rate%· 19 models
  1. #1GPT-5.6 Sol88.0%
  2. #2GPT-5.584.9%
  3. #3Gemini 2.5 Pro83.1%
  4. #4Grok 4.579.6%
  5. #5DeepSeek-V4-Pro74.2%
Official ↗
LiveCodeBench
Continuously updated competitive-programming set
Reasoning
Metric:Pass{'@'}1%· 11 models
  1. #1DeepSeek-V4-Pro100.0%
  2. #2GPT-5.4100.0%
  3. #3GPT-5.4 mini100.0%
  4. #4GPT-5.5100.0%
  5. #5Claude Sonnet 4.6100.0%
Official ↗
EvalPlus
HumanEval+ strict test suite
Coding
Metric:HumanEval+%· 17 models
  1. #1GPT-5.489.0%
  2. #2GPT-5.4 mini89.0%
  3. #3Qwen3.6-Plus87.2%
  4. #4DeepSeek-V4-Pro86.6%
  5. #5DeepSeek-V4-Flash83.5%
Official ↗

LMArenaChat

Metric: Arena Elo·12 models·fetched 2026-07-18

Open official board ↗
#ModelArena EloCross-board ranks
#1
Claude Fable 5Anthropic
1631
A #1
L #2
A—L—E—
#2
Claude Opus 4.8Anthropic
1581
A #4
L #5
A #6
L #7
E #9
#3
Grok 4.5xAI
1558
A #7
L #9
A #4
L—
E #7
#4
Claude Sonnet 4.6Anthropic
1544
A #8
L #10
A #7
L #5
E #6
#5
Kimi K2.6Moonshot AI
1519
A #17
L #16
A #10
L—E—
#6
GPT-5.5OpenAI
1488
A #6
L #3
A #2
L #4
E—
#7
Gemini 3 ProGoogle
1486
A #24
L—A—L—E—
#8
GPT-5.6 SolOpenAI
1486
A #2
L #1
A #1
L—
E #17
#9
GPT-5.4OpenAI
1472
A #9
L #7
A #8
L #2
E #1
#10
Gemini 2.5 ProGoogle
1393
A #31
L—
A #3
L #8
E #8
#11
Gemini 3.5 FlashGoogle
1286
A #12
L #11
A—L—E—
#12
Gemini 3.1 ProGoogle
1211
A #13
L #8
A—L—E—
Rank color:#1#2#3#4-5#6-10#11+

The "Cross-board ranks" column shows this model's rank on the other 5 boards at a glance (grey = the model was not tested on that board).

Hundreds of thousands of human blind-vote conversations (Elo), from LMArena / Chatbot Arena. The most authoritative "user preference" board today.

📊 6 benchmark providers
LMArena ↗ Chat

Hundreds of thousands of human blind-vote conversations (Elo), from LMArena / Chatbot Arena. The most authoritative "user preference" board today.

Arena Elo12 models
Artificial Analysis ↗ Overall

Artificial Analysis aggregates 20+ benchmarks (MMLU-Pro, GPQA, HLE, SciCode, IFBench, Terminal-bench, etc.) into a single Intelligence Index — the most comprehensive commercial evaluator for "overall capability".

AA Index46 models
LiveBench ↗ Reasoning

LiveBench refreshes its questions monthly to prevent contamination. Covers reasoning, math, coding, language, instruction following, and data analysis.

Avg score%23 models
Aider Polyglot ↗ Coding

133 real coding tasks (bug-fix / feature) across 6 languages. Metric: pass_rate_2.

Pass rate%19 models
LiveCodeBench ↗ Reasoning

Continuously updated competitive-programming problems. Metric: Pass@1.

Pass{'@'}1%11 models
EvalPlus ↗ Coding

HumanEval+ strict test suite, 164 tasks.

HumanEval+%17 models

All rankings and numbers come directly from official sources; we only do name matching and aggregate presentation, no weighting or re-ranking. Boards use different methodologies, so do not compare raw values across boards.