BlogLLM Benchmark Scores as Published, Side by Side
ERQA: Results as Published
Every result the catalog holds for ERQA: 21 models from 4 developers, read from 7 pages. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Multimodal UnderstandingMaintainer Google DeepMind (Gemini Robotics team); the repository sits under the embodiedreasoning organisationIntroduced 2025Data version 2026103001, as of 2026-10-30
ERQA is a multimodal benchmark of embodied reasoning questions for robotics-related scenarios. The README says it covers spatial reasoning and world knowledge, and each item mixes images and text in a multiple-choice question with answers labelled A to D. The full benchmark consists of 400 examples. The README does not list named reasoning categories, and its evaluation harness queries the Gemini and OpenAI APIs; the README gives no accuracy formula beyond the ground-truth letter answer per question.
ERQA
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-5.2OpenAI, released 2025-12-11 | 59.8% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15 | - |
| GPT-5OpenAI, released 2025-08-07 | 65.7% | OpenAI, Introducing GPT-5checked 2026-10-15 | Effort: high |
| 42% | OpenAI, Introducing GPT-5checked 2026-10-15 | Effort: none | |
| GPT-5 miniOpenAI, released 2025-08-07 | 54% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| o3OpenAI, released 2025-04-16 | 64% | OpenAI, Introducing GPT-5checked 2026-10-15 | Effort: high |
| GPT-4oOpenAI, released 2024-05-13 | 35.2% | OpenAI, Introducing GPT-5checked 2026-10-15 | - |
| Claude Opus 4.6Anthropic, released 2026-02-05 | 40.8% | Alibaba, Qwen3.8-27B model cardchecked 2026-10-15; printed on 1 more page | Effort: max |
| Claude Opus 4.5Anthropic, released 2025-11-24 | 46.8% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15; printed on 1 more page | - |
| Claude Sonnet 4.5Anthropic, released 2025-09-29 | 45% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| Gemini 4 ArgonGoogle, released 2026-09-30 | 87% | Google, Gemini 4 Argon: Built to solve your hardest problemschecked 2026-10-25 | - |
| Gemma 4 31BGoogle, released 2026-04-02 | 57.5% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15 | - |
| Gemini 3 ProGoogle, released 2025-11-18 | 70.5% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15 | - |
| Qwen3.7-PlusAlibaba, released 2026-06-01 | 69.8% | Alibaba, Qwen3.8-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| Qwen3.6-27BAlibaba, released 2026-04-22 | 62.5% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| Qwen3.6-35B-A3BAlibaba, released 2026-04-16 | 61.8% | Alibaba, Qwen3.6-27B model cardchecked 2026-10-15 | - |
| Qwen3.5-122B-A10BAlibaba, released 2026-02-24 | 62% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| Qwen3.5-27BAlibaba, released 2026-02-24 | 60.5% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page | - |
| Qwen3.5-35B-A3BAlibaba, released 2026-02-24 | 64.8% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15 | - |
| Qwen3.5-397B-A17BAlibaba, released 2026-02-16 | 67.5% | Alibaba, Qwen3.5-397B-A17B model cardchecked 2026-10-15; printed on 1 more page | - |
| Qwen3-VL-235B-A22BAlibaba, released 2025-09-23 | 52.5% | Alibaba, Qwen3.5-122B-A10B model cardchecked 2026-10-15; printed on 1 more page | - |
| Qwen3.8-27BAlibaba | 65.5% | Alibaba, Qwen3.8-27B model cardchecked 2026-10-15; printed on 1 more page | - |
| Qwen3.8-Flash-NextAlibaba | 72.3% | Alibaba, Qwen3.8-Flash-Next model cardchecked 2026-10-15 | - |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.