reasoning benchmark
Arena-Hard v2 Leaderboard
Arena-Hard-Auto v2 is a challenging benchmark consisting of 500 carefully curated prompts sourced from Chatbot Arena and WildChat-1M, designed to evaluate large language models on real-world user queries. The benchmark covers diverse domains including open-ended software engineering problems, mathematics, creative writing, and technical problem-solving. It uses LLM-as-a-Judge for automatic evaluation, achieving 98.6% correlation with human preference rankings while providing 3x higher separation of model performances compared to MT-Bench. The benchmark emphasizes prompt specificity, complexity, and domain knowledge to better distinguish between model capabilities.
Updated Aug 17, 2026
Models16
Model coverage16
MetricScore
EvidenceB
Arena-Hard v2 Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelXI | Score86.2% | Percentile100.0% | Participants16 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score82.7% | Percentile93.3% | Participants16 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score79.7% | Percentile86.7% | Participants16 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score79.2% | Percentile80.0% | Participants16 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score77.4% | Percentile73.3% | Participants16 | EvidenceC | Evaluated |
| Rank06 | ModelNV | Score73.9% | Percentile66.7% | Participants16 | EvidenceC | Evaluated |
| Rank07 | ModelSA | Score71.0% | Percentile60.0% | Participants16 | EvidenceC | Evaluated |
| Rank08 | ModelNV | Score67.7% | Percentile53.3% | Participants16 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score64.7% | Percentile46.7% | Participants16 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score62.3% | Percentile40.0% | Participants16 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score60.5% | Percentile33.3% | Participants16 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score58.5% | Percentile26.7% | Participants16 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score56.7% | Percentile20.0% | Participants16 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score51.1% | Percentile13.3% | Participants16 | EvidenceC | Evaluated |
| Rank15 | ModelSA | Score49.0% | Percentile6.7% | Participants16 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score36.8% | Percentile0.0% | Participants16 | EvidenceC | Evaluated |
Arena-Hard v2 Highlights
The leading models and scores on this benchmark.
Arena-Hard v2 Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for Arena-Hard v2
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis arena-hard v2 AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01XIMiMo-V2-FlashXiaomi
- Score
- 86.2%
- Price
- $0.14 input / $0.28 output per 1M tokens
Strengths
- Ranks #1 of 16 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures Arena-Hard v2, not total model capability
- 02ACQwen3-Next-80B-A3B-InstructAlibaba Cloud / Qwen Team
- Score
- 82.7%
- Price
- $0.50 input / $2.0 output per 1M tokens
Strengths
- Ranks #2 of 16 compared models
- 93th percentile on this benchmark
- C evidence result
Considerations
- This result measures Arena-Hard v2, not total model capability
- 03ACQwen3-235B-A22B-Thinking-2507Alibaba Cloud / Qwen Team
- Score
- 79.7%
Strengths
- Ranks #3 of 16 compared models
- 87th percentile on this benchmark
- C evidence result
Considerations
- This result measures Arena-Hard v2, not total model capability
- 04ACQwen3-235B-A22B-Instruct-2507Alibaba Cloud / Qwen Team
- Score
- 79.2%
- Price
- $0.70 input / $2.8 output per 1M tokens
- Speed
- Up to 68 tok/s via Fireworks
Strengths
- Ranks #4 of 16 compared models
- 80th percentile on this benchmark
- C evidence result
Considerations
- This result measures Arena-Hard v2, not total model capability
- 05ACQwen3 VL 235B A22B InstructAlibaba Cloud / Qwen Team
- Score
- 77.4%
Strengths
- Ranks #5 of 16 compared models
- 73th percentile on this benchmark
- C evidence result
Considerations
- This result measures Arena-Hard v2, not total model capability
Selection summary
Best AI Models for Arena-Hard v2
MiMo-V2-Flash currently leads Arena-Hard v2 with 86.2%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is Arena-Hard v2?
What Arena-Hard v2 measures and how its scores work.
Arena-Hard-Auto v2 is a challenging benchmark consisting of 500 carefully curated prompts sourced from Chatbot Arena and WildChat-1M, designed to evaluate large language models on real-world user queries. The benchmark covers diverse domains including open-ended software engineering problems, mathematics, creative writing, and technical problem-solving. It uses LLM-as-a-Judge for automatic evaluation, achieving 98.6% correlation with human preference rankings while providing 3x higher separation of model performances compared to MT-Bench. The benchmark emphasizes prompt specificity, complexity, and domain knowledge to better distinguish between model capabilities.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- Arena-Hard v2
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- arena-hard-v2|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about Arena-Hard v2.
Which model scores highest on Arena-Hard v2?
MiMo-V2-Flash is currently ranked first with 86.2%.
What does Arena-Hard v2 measure?
Arena-Hard-Auto v2 is a challenging benchmark consisting of 500 carefully curated prompts sourced from Chatbot Arena and WildChat-1M, designed to evaluate large language models on real-world user queries. The benchmark covers diverse domains including open-ended software engineering problems, mathematics, creative writing, and technical problem-solving. It uses LLM-as-a-Judge for automatic evaluation, achieving 98.6% correlation with human preference rankings while providing 3x higher separation of model performances compared to MT-Bench. The benchmark emphasizes prompt specificity, complexity, and domain knowledge to better distinguish between model capabilities.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
16 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.