# We-Math leaderboard

Multimodal & Grounded benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelchoice.dev/leaderboards/benchmarks/multimodal-grounded-we-math

Every model the catalog carries a published We-Math value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | Multimodal & Grounded |
| Measure | Multimodal mathematical reasoning |
| Tasks | Visually grounded math problems |
| Difficulty | Advanced multimodal mathematics |
| Published by | [Qwen3.6 launch benchmarks](https://qwen.ai/blog?id=qwen3.6) |

A multimodal math benchmark for visually grounded mathematical reasoning and answer generation.

## We-Math ranking

6 models with a published We-Math value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.

| Rank | Model | Provider | Multimodal mathematical reasoning | Evidence |
| --- | --- | --- | --- | --- |
| 1 | [Qwen3.5 397B](https://modelchoice.dev/models/qwen3-5-397b) | Alibaba | 87.9 | Observed 2026-11-21 · source benchlm:models |
| 2 | [Gemini 3 Pro](https://modelchoice.dev/models/gemini-3-pro) | Google | 86.9 | Observed 2026-11-21 · source benchlm:models |
| 3 | [Kimi K2.5](https://modelchoice.dev/models/kimi-k2-5) | Moonshot AI | 84.7 | Observed 2026-11-21 · source benchlm:models |
| 4 | [GPT-5.2](https://modelchoice.dev/models/gpt-5-2) | OpenAI | 79.0 | Observed 2026-11-21 · source benchlm:models |
| 5 | [Claude Opus 4.5](https://modelchoice.dev/models/claude-opus-4-5) | Anthropic | 70.0 | Observed 2026-11-21 · source benchlm:models |
| 6 | [LFM3-VL-2B](https://modelchoice.dev/models/lfm3-vl-2b) | LiquidAI | 68.6 | Observed 2026-11-21 · source benchlm:models |

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
