LiveCodeBench (AA)
As of 29 Oct 2026, the top score on LiveCodeBench (AA) is 91.8% by Gemini 3 Pro (Google), ahead of Gemini 3 Flash at 90.8%, across 304 model configurations with a published result (Artificial Analysis, result dated 18 Nov 2025). Competitive programming problems harvested continuously from LeetCode, AtCoder and Codeforces after a model's training cut-off, so it is hard to have memorised. Run by Artificial Analysis. LiveCodeBench (AA) is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 29 Oct 2026)
- Gemini 3 Pro holds the top LiveCodeBench (AA) score at 91.8% (Artificial Analysis, 18 Nov 2025).
- Gemini 3 Flash is second on LiveCodeBench (AA) at 90.8% (Artificial Analysis, 17 Dec 2025).
- LiveCodeBench (AA): Competitive programming problems harvested continuously from LeetCode, AtCoder and Codeforces after a model's training cut-off, so it is hard to have memorised. Run by Artificial Analysis (Artificial Analysis).
- 304 model configurations have a published LiveCodeBench (AA) result on BenchLeader (BenchLeader, 29 Oct 2026).
- LiveCodeBench (AA) does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Competitive programming problems harvested continuously from LeetCode, AtCoder and Codeforces after a model's training cut-off, so it is hard to have memorised. Run by Artificial Analysis.
- Published by
- Artificial Analysis
- Category
- Coding
- Index weight
- Reference only
- Models
- 304
- Data as of
- 29 Oct 2026
- Newest result
- 4 Jan 2026
The publisher has posted no new results on this board since 4 Jan 2026. BenchLeader re-reads it every day, so newer models appear here as soon as they are published.
Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.
- 1Gemini 3 Pro (high)91.8%
- 2Gemini 3 Flash (thinking)90.8%
- 3DeepSeek V3.2 Speciale89.6%
- 4GPT-5.2 (medium)89.4%
- 5GLM 4.7 (thinking)89.4%
- 6GPT-5.2 (xhigh)88.9%
- 7gpt-oss-120b (high)87.8%
- 8Claude Opus 4.5 (thinking)87.1%
- 9GPT-5.1 (high)86.8%
- 10MiMo-V2-Flash (thinking)86.8%
- 11DeepSeek V3.2 (thinking)86.2%
- 12o4-mini (high)85.9%
- 13Gemini 3 Pro (low)85.7%
- 14Kimi K2 Thinking (thinking)85.3%
- 15GPT-5.1-Codex (high)84.9%
304 of 304
| # | ||||
|---|---|---|---|---|
| 1 |  Gemini 3 Prohigh | 91.8% | 60.1 | 2025-11-18 |
| 2 | Gemini 3 Flashthinking | 90.8% | 59.0 | 2025-12-17 |
| 3 |  DeepSeek V3.2 Specialeopen ↗ | 89.6% | 53.6 | 2025-12-01 |
| 4 |  GPT-5.2medium | 89.4% | 56.2 | 2025-12-11 |
| 5 |  GLM 4.7thinkingopen ↗ | 89.4% | – | 2025-12-22 |
| 6 | GPT-5.2xhigh | 88.9% | 60.1 | 2025-12-11 |
| 7 | gpt-oss-120bhighopen ↗ | 87.8% | 49.8 | 2025-08-05 |
| 8 |  Claude Opus 4.5thinking | 87.1% | 59.6 | 2025-11-24 |
| 9 | GPT-5.1high | 86.8% | 57.7 | 2025-11-13 |
| 10 |  MiMo-V2-Flashthinkingopen ↗ | 86.8% | 54.0 | 2025-12-16 |
| 11 | DeepSeek V3.2thinkingopen ↗ | 86.2% | 56.7 | 2025-12-01 |
| 12 | o4-minihigh | 85.9% | 50.3 | 2025-04-16 |
| 13 | Gemini 3 Prolow | 85.7% | – | 2025-11-18 |
| 14 |  Kimi K2 Thinkingthinkingopen ↗ | 85.3% | 53.3 | 2025-11-06 |
| 15 | GPT-5.1-Codexhigh | 84.9% | – | 2025-11-13 |
| 16 | GPT-5high | 84.5% | 57.9 | 2025-08-07 |
| 17 | GPT-5-Codexhigh | 84.0% | 58.6 | 2025-09-23 |
| 18 | GPT-5 minihigh | 83.8% | 54.6 | 2025-08-07 |
| 19 | GPT-5.1-Codex-Minihigh | 83.6% | – | 2025-11-13 |
| 20 |  Grok 4 Fastthinking | 83.2% | 51.7 | 2025-09-19 |
| 21 |  MiniMax-M2open ↗ | 82.7% | 52.6 | 2025-10-26 |
| 22 | Grok 4.1 Fastthinking | 82.2% | 55.6 | 2025-11-19 |
| 23 | Grok 4 | 81.9% | 58.4 | 2025-07-10 |
| 24 |  ERNIE 5.0thinking | 81.2% | – | 2025-11-13 |
| 25 | MiniMax-M2.1open ↗ | 81.0% | 55.0 | 2025-12-23 |
| 26 | o3 | 80.8% | 61.0 | 2025-04-16 |
| 27 |  Apriel-v1.6-15B-Thinkeropen | 80.7% | – | 2025-11-25 |
| 28 | Gemini 2.5 Pro | 80.1% | 54.5 | 2025-06-05 |
| 29 | DeepSeek V3.1 Terminusthinkingopen ↗ | 79.8% | 55.6 | 2025-09-22 |
| 30 | Gemini 3 Flashno reasoning | 79.7% | – | 2025-12-17 |
| 31 | GPT-5 nanohigh | 78.9% | 47.4 | 2025-08-07 |
| 32 |  Qwen3 235B A22B 2507thinkingopen ↗ | 78.8% | 51.9 | 2025-07-25 |
| 33 | Qwen3-Next 80B A3Bthinkingopen ↗ | 78.4% | 52.3 | 2025-09-11 |
| 34 | DeepSeek V3.1thinkingopen ↗ | 78.4% | – | 2025-08-21 |
| 35 |  Intellect 3open ↗ | 77.7% | 48.6 | 2025-11-27 |
| 36 | gpt-oss-20bhighopen ↗ | 77.7% | 45.6 | 2025-08-05 |
| 37 | DeepSeek R1 0528open ↗ | 77.0% | 49.4 | 2025-05-28 |
| 38 |  K-EXAONEthinkingopen | 76.8% | – | 2025-12-31 |
| 39 | Qwen3 Maxmax | 76.7% | 54.3 | 2025-09-23 |
| 40 |  Doubao-Seed-Code | 76.6% | – | 2025-11-11 |
| 41 | Seed-OSS-36B-Instructopen ↗ | 76.5% | 49.2 | 2025-08-20 |
| 42 | GPT-5low | 76.3% | 51.4 | 2025-08-07 |
| 43 | GPT-5 nanomedium | 76.3% | 45.5 | 2025-08-07 |
| 44 |  Magistral Medium 1.2 | 75.0% | 43.8 | 2025-09-18 |
| 45 |  Kat Coder Pro v1 | 74.7% | – | 2025-11-11 |
| 46 | EXAONE 4.0 32Bthinkingopen ↗ | 74.7% | – | 2025-07-15 |
| 47 |  Nemotron 3 Nano 30B A3Bthinkingopen ↗ | 74.1% | 47.5 | 2025-12-15 |
| 48 | Claude Opus 4.5no reasoning | 73.8% | 52.1 | 2025-11-24 |
| 49 | GLM-4.5thinkingopen ↗ | 73.8% | – | 2025-07-28 |
| 50 | Qwen3 VL 32B Instructthinkingopen ↗ | 73.8% | – | 2025-10-21 |
| 51 | Llama Nemotron Super 49B v1.5thinkingopen | 73.7% | – | 2025-07-25 |
| 52 | o3-minihigh | 73.4% | 47.7 | 2025-01-31 |
| 53 |  Nova 2.0 Pro Previewmedium | 73.0% | 53.6 | 2025-11-27 |
| 54 |  Falcon-H1R-7Bopen ↗ | 72.4% | – | 2026-01-04 |
| 55 | NVIDIA Nemotron Nano 9B v2thinkingopen | 72.4% | – | 2025-08-18 |
| 56 | Magistral Small 1.2open | 72.3% | 40.0 | 2025-09-17 |
| 57 | o3-mini | 71.8% | 50.7 | 2025-01-31 |
| 58 | Claude Sonnet 4.5thinking | 71.4% | 52.3 | 2025-09-29 |
| 59 | Gemini 2.5 Flash 09 2025thinking | 71.3% | 52.9 | 2025-09-25 |
| 60 | MiniMax-M1open | 71.1% | 49.7 | 2025-06-17 |
More coding benchmarks · All LLM benchmarks
Related benchmarks
Cite as: BenchLeader, “LiveCodeBench (AA) leaderboard”, data as of 29 Oct 2026, https://www.benchleader.com/benchmarks/aa_livecodebench.
LiveCodeBench (AA): questions
- What does LiveCodeBench (AA) measure?
- Competitive programming problems harvested continuously from LeetCode, AtCoder and Codeforces after a model's training cut-off, so it is hard to have memorised. Run by Artificial Analysis. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on LiveCodeBench (AA)?
- As of 29 Oct 2026, the highest score on LiveCodeBench (AA) is 91.8% by Gemini 3 Pro (result dated 18 Nov 2025), ahead of Gemini 3 Flash at 90.8%. Scores come from Artificial Analysis and are re-read daily.
- Which AI model leads LiveCodeBench (AA)?
- Gemini 3 Pro leads LiveCodeBench (AA) with 91.8% as of 29 Oct 2026, ahead of Gemini 3 Flash at 90.8% (result dated 18 Nov 2025).
- How many models have LiveCodeBench (AA) results?
- 304 model configurations have a LiveCodeBench (AA) result on BenchLeader, all taken from Artificial Analysis.
- Who runs LiveCodeBench (AA) and how often is it updated?
- LiveCodeBench (AA) is published by Artificial Analysis. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does LiveCodeBench (AA) count toward the BenchLeader Index?
- No. LiveCodeBench (AA) is shown for reference but left out of the composite index.