Best Multimodal LLMs in 2026
As of October 10, 2026, the top model among multimodal LLMs on the BenchLM leaderboard is Kimi K3 with a score of 92.6.
Bottom line: the live table ranks how well models read images, documents, and charts. For image generation, look at dedicated image models — this table measures visual understanding.
This ranking moves when models ship. Get the releases, price changes and retirements that affect your shortlist. Follow model changes
Full Rankings (53 ranked, 1 tracked)
Public preview through Mistral Studio; Mistral says the weights are due by the end of October 2026.
—
tracked
Current position
Kimi K3 leads the live multimodal llms ranking at 92.6.
Claude Opus 5 ranks #2 at 92.2.
Claude Opus 5.5 ranks #3 at 92.2.
How to choose
Key Takeaways
The top model is Kimi K3 by Moonshot AI with a provisional score of 92.6.
The best open-weight model is Qwen3.8 Max at position #4.
53 models are ranked here; 1 tracked model is listed without a position.
Score in Context
What these scores mean
The multimodal & grounded score blends image understanding, document/chart reading, and visually grounded reasoning benchmarks. Higher means the model more reliably extracts and reasons over visual content.
Known limitations
These are understanding scores — they say nothing about image generation quality. OCR-heavy production workloads should also test resolution limits and page-count behavior, which benchmarks only partially cover.
About this ranking
Ranking data as of October 10, 2026
This page ranks models by BenchLM's multimodal & grounded category: image understanding, document and chart reading, and visually grounded reasoning. Note the scope: these scores measure how well a model understands images — not how well it generates them. Image-generation systems are a different product class and are not ranked here.
Unless noted otherwise, ranking surfaces on this page use BenchLM’s provisional leaderboard lane rather than the stricter sourced-only verified leaderboard.
Kimi K3 leads this ranking with a score of 92.6, followed by Claude Opus 5 (92.2) and Claude Opus 5.5 (92.2). The top three are separated by just a few points — any of them would perform well for this use case.
The best open-weight option is Qwen3.8 Max (ranked #4 with a score of 91.2). While proprietary models lead, open-weight options are within striking distance for teams willing to trade a few points of performance for full model control.
This ranking uses provisional weighted averages across the scoring benchmarks in multimodalGrounded. For detailed model profiles, click any model name above. To compare two specific models head-to-head, use the "vs #" links.
Questions
What is the best multimodal LLM?
The first row of the multimodal & grounded table above is the current answer. The table recomputes with every data refresh, so check the date on the ranking. For a budget pick, check the multimodal value ranking, which divides this category score by output price, then test your own images and documents.
What is the best LLM for image generation?
This page ranks image understanding, not generation — LLM leaderboards measure how well models read and reason over images. Image generation is a separate product class (diffusion and autoregressive image models) that BenchLM's categories do not currently score. Treat any "LLM" ranking of image generation with skepticism.
Which LLM is best for reading documents and PDFs?
Document understanding tracks the multimodal & grounded category closely, so start from the top rows of the table above. Long-context support matters as much as vision quality for multi-hundred-page work — see the large-context rankings for models pairing 1M-token windows with strong vision.
Can these models understand charts and tables?
The top rows handle standard charts and tables well — chart-reading benchmarks feed this category. Reliability drops on dense dashboards, small text, and unusual chart types, so test your hardest real documents rather than trusting a single aggregate score.
Explore More
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 5,500+ readers.
One email each week. Unsubscribe anytime.