OmniDocBench 1.5
We show this table for reference; we do not rank on it.
A document understanding benchmark used in frontier-model comparison tables to measure extraction and grounded reasoning quality on complex documents.
Benchmark score on OmniDocBench 1.5 — November 13, 2026
We compile the OmniDocBench 1.5 rows from provider self-reports and secondary reports. Qwen3.8 Max leads the table at 92.1%, followed by MiniMax M3 (91.6%) and Qwen3.7 Plus (91.4%). We do not use these results to rank models overall.
Qwen3.8 Max
Alibaba
Reported by AlibabaMiniMax M3
MiniMax
Reported by MiniMaxQwen3.7 Plus
Alibaba
Reported by Alibaba11 modelsMultimodal & GroundedCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (11 models)
None of the 11 results on this page has a documented run by the benchmark or an independent evaluator: 8 reported by the labs that built the models and 3 reported in another publisher’s document.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Qwen3.8 MaxAlibabaReported by Alibaba | 92.1% | Reported by Alibaba | Not reported | Open |
| 2 | MiniMax M3MiniMaxReported by MiniMax | 91.6% | Reported by MiniMax | Not reported | Open |
| 3 | Qwen3.7 PlusAlibabaReported by Alibaba | 91.4% | Reported by Alibaba | Not reported | Closed |
| 4 | Qwen3.8-27BAlibabaReported by Alibaba | 91.1% | Reported by Alibaba | Not reported | Open |
| 5 | Qwen3.5 397BAlibabaReported by Alibaba | 90.8% | Reported by Alibaba | Not reported | Open |
| 6 | Qwen3.6-35B-A3BAlibabaReported by Alibaba | 89.9% | Reported by Alibaba | Not reported | Open |
| 7 | Kimi K2.5Moonshot AIReported by Moonshot AI | 88.8% | Reported by Moonshot AI | Not reported | Open |
| 8 | Gemini 3 ProGoogleAnother lab’s comparison table | 88.5% | Another lab’s comparison table | Not reported | Closed |
| 9 | Claude Opus 4.5AnthropicAnother lab’s comparison table | 87.7% | Another lab’s comparison table | Not reported | Closed |
| 10 | GPT-5.2OpenAIAnother lab’s comparison table | 85.7% | Another lab’s comparison table | Not reported | Closed |
| 11 | Muse Glimmer 30BMetaReported by Meta | 75.8% | Reported by Meta | Not reported | Open |
- Lab-reported
- Benchmark-run
- Independent
Among the reported OmniDocBench 1.5 rows, Qwen3.8 Max is first at 92.1%. The third row is 0.7 points behind. The broader top-10 range is 6.4 points, so many of the published results sit in a relatively narrow band.
11 models have been evaluated on OmniDocBench 1.5. The benchmark falls in the Multimodal & Grounded category. OmniDocBench 1.5 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About OmniDocBench 1.5
Year
2026
Tasks
Document understanding tasks
Format
Document understanding benchmark
Difficulty
Grounded document reasoning
BenchLM stores OmniDocBench 1.5 as the higher-is-better score format used in current first-party comparison tables. Earlier low-is-better error-style rows are intentionally not mixed into this benchmark key.
Freshness and provenance
Version
OmniDocBench 1.5 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does OmniDocBench 1.5 measure?
A document understanding benchmark used in frontier-model comparison tables to measure extraction and grounded reasoning quality on complex documents.
Which model scores highest on OmniDocBench 1.5?
Qwen3.8 Max by Alibaba currently leads with a score of 92.1% on OmniDocBench 1.5.
How many models are evaluated on OmniDocBench 1.5?
11 AI models have published results on OmniDocBench 1.5 in the BenchLM catalog.
Compare top models on OmniDocBench 1.5
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.