Benchmark
CharXiv-R Benchmark
What is CharXiv-R?
CharXiv-R is the reasoning component of the CharXiv benchmark, focusing on complex reasoning questions that require synthesizing information across visual…
- Modality
- multimodal
- Evaluates
- multimodal, reasoning, vision
- Openness
- unknown
- Importer
- LLM Stats
- Reported scores
- 51
Source provenance
- Original evidence https://api.zeroeval.com/leaderboard/benchmarks/charxiv-r?top_n=500
CharXiv-R code
CharXiv-R results and reported scores
LLM Stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Claude Mythos Preview | Anthropic | 0.932 | 2026-04-07 | evidence |
| Claude Opus 4.6 | Anthropic | 0.774 | 2026-02-05 | evidence |
| Claude Opus 4.7 | Anthropic | 0.91 | 2026-04-16 | evidence |
| Claude Opus 4.8 | Anthropic | 0.899 | 2026-05-28 | evidence |
| Claude Sonnet 5 | Anthropic | 0.883 | 2026-06-30 | evidence |
| Command A+ | Cohere | 0.527 | 2026-05-20 | evidence |
| Gemini 3 Flash | 0.803 | 2025-12-17 | evidence | |
| Gemini 3 Pro | 0.814 | 2025-11-18 | evidence | |
| Gemini 3.1 Flash-Lite | 0.732 | 2026-03-03 | evidence | |
| Gemini 3.5 Flash | 0.842 | 2026-05-19 | evidence | |
| Gemini 3.5 Flash-Lite | 0.765 | 2026-07-21 | evidence | |
| Gemini 3.6 Flash | 0.894 | 2026-07-21 | evidence | |
| Gemini 3.7 Flash | 0.887 | 2026-08-13 | evidence | |
| GPT-4.1 | OpenAI | 0.567 | 2025-04-14 | evidence |
| GPT-4.1 mini | OpenAI | 0.568 | 2025-04-14 | evidence |
| GPT-4.1 nano | OpenAI | 0.405 | 2025-04-14 | evidence |
| GPT-4.5 | OpenAI | 0.554 | 2025-02-27 | evidence |
| GPT-4o | OpenAI | 0.588 | 2024-08-06 | evidence |
| GPT-5 | OpenAI | 0.811 | 2025-08-07 | evidence |
| GPT-5.2 | OpenAI | 0.821 | 2025-12-11 | evidence |
| GPT-5.5 Instant | OpenAI | 0.816 | 2026-05-05 | evidence |
| Inkling-Small | Thinking Machines Lab | 0.774 | 2026-07-30 | evidence |
| Kimi K2.5 | Moonshot AI | 0.775 | 2026-01-27 | evidence |
| Kimi K2.6 | Moonshot AI | 0.867 | 2026-04-20 | evidence |
| Kimi K3 | Moonshot AI | 0.913 | 2026-07-16 | evidence |
| MiMo-V2.5 | Xiaomi | 0.81 | 2026-04-22 | evidence |
| Muse Glimmer-30B | Meta | 0.788 | 2026-08-10 | evidence |
| Muse Spark | Meta | 0.864 | 2026-04-08 | evidence |
| Muse Spark 1.1 | Meta | 0.884 | 2026-07-09 | evidence |
| o3 | OpenAI | 0.786 | 2025-04-16 | evidence |
| o4-mini | OpenAI | 0.72 | 2025-04-16 | evidence |
| Qwen3 VL 235B A22B Instruct | Qwen | 0.621 | 2025-09-22 | evidence |
| Qwen3 VL 235B A22B Thinking | Qwen | 0.661 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Instruct | Qwen | 0.489 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Thinking | Qwen | 0.566 | 2025-09-22 | evidence |
| Qwen3 VL 32B Instruct | Qwen | 0.628 | 2025-09-22 | evidence |
| Qwen3 VL 32B Thinking | Qwen | 0.652 | 2025-09-22 | evidence |
| Qwen3 VL 4B Instruct | Qwen | 0.397 | 2025-09-22 | evidence |
| Qwen3 VL 4B Thinking | Qwen | 0.503 | 2025-09-22 | evidence |
| Qwen3 VL 8B Instruct | Qwen | 0.464 | 2025-09-22 | evidence |
| Qwen3 VL 8B Thinking | Qwen | 0.53 | 2025-09-22 | evidence |
| Qwen3.5-122B-A10B | Qwen | 0.772 | 2026-02-24 | evidence |
| Qwen3.5-27B | Qwen | 0.795 | 2026-02-24 | evidence |
| Qwen3.5-35B-A3B | Qwen | 0.775 | 2026-02-24 | evidence |
| Qwen3.6 Plus | Qwen | 0.815 | 2026-04-02 | evidence |
| Qwen3.6-27B | Qwen | 0.784 | 2026-04-21 | evidence |
| Qwen3.6-35B-A3B | Qwen | 0.78 | 2026-04-16 | evidence |
| Qwen3.7-Plus | Qwen | 0.859 | 2026-05-31 | evidence |
| Qwen3.8-27B | Qwen | 0.902 | 2026-08-14 | evidence |
| Seed 2.1 Pro | ByteDance | 0.864 | 2026-06-24 | evidence |
| Seed 2.1 Turbo | ByteDance | 0.836 | 2026-06-24 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.