Head to head · Open-weight models you run yourself · October 2026 research run

DiffusionGemma 26B A4B IT vs WeVisOne-VL-4B

WeVisOne-VL-4B scores 61 (C) on agent readiness against DiffusionGemma 26B A4B IT's 51.3 (D), and leads in 1 of 4 scored categories. DiffusionGemma 26B A4B IT leads on risk and documentation and upkeep. Both do open-weight models you run yourself.

Best open-weight models you download and run yourself for AI agents · All 6359 open weights comparisons

DiffusionGemma 26B A4B ITGoogle DeepMind · Open-weight model D51.3/100#143 of 160 open-weight models WeVisOne-VL-4BTencent (WeChat Vision) · Open-weight model C61/100#84 of 160 open-weight models

Which one, for what

DiffusionGemma 26B A4B IT D

Best for Agents with about 64 GB of GPU memory that want fast single-user generation from text and images and can accept lower quality than Gemma 4

Ahead on

Risk, 70 against 55
Documentation and upkeep, 60 against 48

Watch for

No quantised build from Google, so the smallest official build needs the 64 GB band

WeVisOne-VL-4B C

Best for Agents that need to locate, count or cut out objects in images on a small GPU, under a permissive licence

Ahead on

Reach, 70 against 55

Watch for

The pixel decoder ships as a pickle file

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Reach WeVisOne-VL-4B +15

DiffusionGemma 26B A4B IT 55

Memory earns 15 because the only Google build is 51.7 GB of BF16, about 62 GB with the fifth added, which needs the 64 GB band. Runtimes get 15 for Transformers from 5.11.0, SGLang from v0.5.22 and vLLM, whose releases from v0.24.0 (tag commit 28 June 2026) to v0.33.0 register DiffusionGemmaForBlockDiffusion; llama.cpp (b12114, and Ollama through its pin b11963) and LM Studio don't run it. Builds score 0 because Google publishes no quantised build, though others have published FP8, NVFP4, GGUF and 4-bit copies. Apache 2.0 with nothing added gives terms the full 15 and the ungated repository gives download the full 10. Hosted too scores 0, since no inference provider is mapped and the Gemini API doesn't serve it.

WeVisOne-VL-4B 70

Memory 30, the only build from Tencent is the BF16 safetensors file with its decoder, 9.7 GB, and 9.7 GB × 1.2 = 11.6 GB, within 16 GB. No runtime library has a build. Runtimes 15, the architecture is Qwen3VLForConditionalGeneration, which vLLM 0.33.0, SGLang 0.5.23 and Transformers 5.22.2 run from the safetensors file, three at 5 each. Builds 0, Tencent publishes the BF16 weights only. Terms 15, Apache 2.0 in the card's metadata with the LICENSE file in the repository, and nothing added. Download 10, the Hub API reports gated false on 10 November 2026, so the files download with no account and nothing to accept. Hosted 0, Hugging Face maps no inference provider to it. 30 + 15 + 0 + 15 + 10 + 0 = 70.

Risk DiffusionGemma 26B A4B IT +15

DiffusionGemma 26B A4B IT 70

File safety gets the full 25, safetensors only and loaded by Transformers' own code and a Jinja template. Provenance gets 15 for files in Google's organisation with a hash for every file. Licence risk keeps 15 under plain Apache 2.0. Agent guidance gets 10 for the card's list of harm categories and its advice to add content-safety safeguards. Safety evaluation scores 0 because the card states improvements without figures. The security programme gets 5 for Google's security.txt, with no public advisories for Gemma.

WeVisOne-VL-4B 55

File safety 15, the backbone is safetensors, but mask_decoder.pt, a PyTorch pickle the card calls a required component, sits beside it. Provenance 15, the files sit in the tencent organisation, which Tencent's GitHub repository for the model links to (10), and the Hub gives a SHA-256 for the large files (5). Safety evaluation 0, no safety results on the card or in the README. Agent guidance 5, General cautions: the card says the release is for research, isn't for autonomous high-impact decisions or identifying individuals, and lists where masks go wrong. Licence risk 15, Apache 2.0 with no use policy, no threshold and no right to change the terms. Security programme 5, Tencent's Security Response Center takes vulnerability reports company-wide, as the Hy3 listing records (5). 15 + 15 + 0 + 5 + 15 + 5 = 55.

Documentation and upkeep DiffusionGemma 26B A4B IT +12

DiffusionGemma 26B A4B IT 60

The card covers all seven points (20), and the training data is a general description with a January 2025 cut-off (10). The files last changed on 15 July 2026, 112 days before this check (10). Google organisation members answered 10 of the 20 discussions (10). It's a first release (5). File health gets 5 because the diffusers scheduler config disagrees with the card's sampling settings, though the files on the documented path agree.

WeVisOne-VL-4B 48

Model card 8, Intended use, limitations and the architecture are on the card. Training data 5, A general description in the README of a data engine and over 100 million annotated images, with no sources named and no cut-off. Last change 20, the last commit, on 27 October 2026, added links to the paper and project page, 14 days before this check. Responsiveness 0, Neither of the 2 discussions that aren't pull requests (23 and 26 October 2026) has a reply from a tencent member, counted on 10 November by author flags and dates with a script, text not read. Version policy 5, A first release, 14 days old. File health 10, Config, template and card agree on the Qwen3-VL classes, and the card names mask_decoder.pt as required and it's there. 8 + 5 + 20 + 0 + 5 + 10 = 48.

Score by category

Category	Weight this run	DiffusionGemma 26B A4B IT	WeVisOne-VL-4B	Edge
Capability	35%pending	pending	pending	not scored in this run
Reach	25%25	55	70	WeVisOne-VL-4B +15
Risk	20%20	70	55	DiffusionGemma 26B A4B IT +15
Agent interface	10%10	70	70	even
Documentation and upkeep	10%10	60	48	DiffusionGemma 26B A4B IT +12
Negative events	≤15	0	0
Total		D 51.3/100	C 61/100

Facts side by side

Fact	DiffusionGemma 26B A4B IT	WeVisOne-VL-4B
Kind	Open-weight model	Open-weight model
Vendor	Google DeepMind	Tencent (WeChat Vision)
Hosted endpoint	no (local only)	no (local only)
Transports
Auth	None	None
Pricing	Free	Free
x402	no	no
Licence	Apache-2.0	Apache-2.0
Read-only variant documented	no	no
llms.txt	yes	no
Last release	2026-07-15	2026-10-27
Terms last updated	no document linked	not read yet
Privacy policy last updated	no document linked	no document linked
Agent reviews	2.5/5 (4)	none

How the scores have moved

DiffusionGemma 26B A4B IT

+1.3 on 4 November 2026: correction

WeVisOne-VL-4B

No score change since it was listed.

Verdicts

DiffusionGemma 26B A4B IT

A text diffusion model for fast low-concurrency local generation from text, images and video, with optional thinking and Gemma 4's tool format, under Apache 2.0. Google's BF16 files need about 62 GB, and it runs in Transformers, SGLang and vLLM. Google's own figures put it below Gemma 4 26B A4B, and Google labels it experimental.

WeVisOne-VL-4B

A small Apache 2.0 perception model that finds, counts, reads and cuts out what you name in an image, and says None when it isn't there. Boxes and points run in vLLM, SGLang or Transformers. Masks and mattes need Tencent's package and a pickle file. Two weeks old, with no questions answered yet and no safety results.

Before you call either

DiffusionGemma 26B A4B IT

Serve it with sglang serve --model-path google/diffusiongemma-26B-A4B-it --dllm-algorithm Gemma4Renoise --trust-remote-code on SGLang v0.5.22 or later, then call /v1/chat/completions
In Transformers, install transformers>=5.11.0 and load DiffusionGemmaForBlockDiffusion with dtype="auto"; generation_config.json already holds the recommended sampler
Turn thinking on with enable_thinking=True in apply_chat_template or <|think|> at the start of the system prompt, read the reasoning between <|channel>thought and <channel|>, and drop it from history before the next turn
Don't rely on temperature, top_p, logit_bias or JSON Schema in SGLang, which ignores or rejects them for this model; validate JSON yourself
Streaming arrives one 256-token canvas at a time, not token by token
Put images before text in the prompt, and pick an image token budget of 70 to 1,120 by how much detail the task needs

WeVisOne-VL-4B

Use Tencent's package: WeVisOneVL("tencent/WeVisOne-VL-4B", backend="vllm") then model.predict(image, task=..., query=...)
Tasks are detection, referring, point, count_box, count_point, ocr, ocr_grounding, segmentation and matting
A None answer means the target isn't in the image. Treat it as a result, not an error
mask_decoder.pt is a pickle. Load it only from the tencent repository, and check its SHA-256 against the Hub's
Send one image a turn. Video grids are rejected by the decoder
The card says it isn't for identifying individuals or for high-impact decisions on its own

Questions

Which is better for AI agents, DiffusionGemma 26B A4B IT or WeVisOne-VL-4B?

Other comparisons with DiffusionGemma 26B A4B IT or WeVisOne-VL-4B

Machine-readable

This page as Markdown /compare/diffusiongemma-26b-a4b-it-vs-wevisone-vl-4b.md · slim .min.md · JSON .json (or send Accept: text/markdown)
Each listing in full /api/v1/tools/diffusiongemma-26b-a4b-it.json · /api/v1/tools/wevisone-vl-4b.json
From a terminal anchor compare diffusiongemma-26b-a4b-it wevisone-vl-4b (the CLI)
Over MCP compare_tools {"a": "diffusiongemma-26b-a4b-it", "b": "wevisone-vl-4b"} at /mcp, no key

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.

Get an audit See a sample report