WeVisOne-VL-4B
- In Input modalities: TextImage
- Open weights
Prices
0 offerings · prices as listedNo priced offering is listed for this model yet.
Details
- Licence
- apache-2.0
- Family
- wevisone-vl
Not stated on Tencent’s pages: max input, output modalities, reasoning, reasoning efforts, reasoning can be off, max thinking budget, tool use, structured output, built-in tools, knowledge cutoff, aliases.
Run locally
Weights: tencent/WeVisOne-VL-4B- Parameters 4.8B
- Dense
- Native precision BF16
- Layers 36
- KV cache per token 144 KB
- Config positions 256K (config.json max_position_embeddings)
- Needs
- ≈ 10.6 GB (weights 9.0 (maker’s file) + KV cache 1.1 + overhead 0.5)
- Usable
- 21.6 GB (90% of its memory)
- Generation speed
- 46–70 tok/s
- Longest context
- 86K (longer doesn’t fit)
- First token
- about 0.16–0.23 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 9.0 GB | ≈ 10.6 | ●Runs well 46–70 tok/s | 86K |
| Q8_0 | 4.8 GB | ≈ 6.4 | ●Runs well 79–119 tok/s | 116K |
| FP8 | 4.5 GB | ≈ 6.1 | ●Runs well 83–125 tok/s | 118K |
| Q6_K | 3.7 GB | ≈ 5.3 | ●Runs well 97–146 tok/s | 124K |
| Q5_K_M | 3.2 GB | ≈ 4.8 | ●Runs well 109–163 tok/s | 127K |
| Q5_0 | 3.1 GB | ≈ 4.7 | ●Runs well 111–167 tok/s | 128K |
| Q4_K_M | 2.7 GB | ≈ 4.4 | ●Runs well 122–182 tok/s | 130K |
| Q4_0 | 2.5 GB | ≈ 4.2 | ●Runs well 128–193 tok/s | 132K |
| IQ4_XS | 2.4 GB | ≈ 4.0 | ●Runs well 134–200 tok/s | 133K |
| Q3_K_Mquality drops | 2.2 GB | ≈ 3.8 | ●Runs well 141–212 tok/s | 134K |
| IQ3_XXSquality drops | 1.7 GB | ≈ 3.3 | ●Runs well 165–247 tok/s | 137K |
| Q2_Kquality drops | 1.7 GB | ≈ 3.3 | ●Runs well 167–250 tok/s | 138K |
| IQ2_XXSquality drops | 1.2 GB | ≈ 2.8 | ●Runs well 206–308 tok/s | 141K |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU GeForce RTX 3080; unified memory Apple M5 · 16 GB.
Run commands
vllm serve tencent/WeVisOne-VL-4B --max-model-len 8192
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by usThis model isn’t on any leaderboard we cite yet.
Changes
All changes →- New language model
Sources
- tencent/WeVisOne-VL-4B config.json (opens in a new tab) huggingface.co · checked Oct 30, 2026
- tencent/WeVisOne-VL-4B model card (opens in a new tab) huggingface.co · checked Oct 30, 2026
- tencent/WeVisOne-VL-4B model card (opens in a new tab) huggingface.co · checked Nov 4, 2026
- tencent/WeVisOne-VL-4B files (opens in a new tab) huggingface.co · checked Nov 5, 2026
Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data
Tencent · China · tencent.com
Related models
Snapshots, the same family and other models from this provider