Up to date
Qwen Apache 2.0 out 12 months ago
4.44B parameters, images and text in, text out, 256K tokens of context. 1 provider serves it.
Verified 2026-10-25 Rank #75 in LLMs · score 103.7 method v1 Sources 3 Downloads 30d 3.8M · #38 of 619 tracked
Changed this week: 6 new quantized builds on the Hub (114 in all).2026-10-24 · 5 new quantized builds on the Hub (113 in all).2026-10-24
Picked by rule from 18 tracked pieces: the lab's own post, the most shared, a field report, the latest.
OpenAI released gpt-oss-120b and gpt-oss-20b, their first open-weight models since GPT-2 in 2019.
The 3 picks above, plus these 15. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
| # | Cheapest live row | In /1M | Out /1M | Context | Quant | Seen |
|---|---|---|---|---|---|---|
| 1 | Featherless | — | — | — | — | Oct 25 |
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Served by 1 API provider; heavily run locally (114 quantized builds); a big base for builders (732 fine-tunes, adapters and merges).
Top Spaces running it: prithivMLmods/Qwen3-VL-Outpost ♥ 208 · embedl/hfviewer ♥ 181 · artificialguybr/qwen-vl ♥ 126 · sergiopaniego/vlm_object_understanding ♥ 126
Weights only, an estimate: leave room for the KV cache. Method
About 2.7 GB at INT4, 5.3 GB at FP8, 11 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2025-10-11), amber triangles are 4 dated articles (each links to it), grey ticks are repo commits.
846 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Listed in LLMs
| Pipeline tag | image-text-to-text |
|---|---|
| Library | transformers |
| Architectures | Qwen3VLForConditionalGeneration |
| Model type | qwen3_vl |
| Dtype | bfloat16 |
| Params by dtype | BF16: 4,437,815,808 |
| Size band | M |
| MoE | no |
| Context length | 262,144 (config.json max_position_embeddings) |
| Papers | arXiv:2505.09388, arXiv:2502.13923, arXiv:2409.12191, arXiv:2308.12966 |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Qwen3VLForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3VLForConditionalGeneration · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Weights-only memory for Qwen3-VL-4B-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 11 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 5.3 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 2.7 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 114finetune: 503adapter: 227merge: 2
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| lmstudio-community/Qwen3-VL-4B-Instruct-MLX-4bit | quantized | MLX | 4.44B | M | apache-2.0 | lmstudio-community | 12 | 2.7M |
| lmstudio-community/Qwen3-VL-4B-Instruct-MLX-8bit | quantized | MLX | 4.44B | M | apache-2.0 | lmstudio-community | 2 | 2.6M |
| lmstudio-community/Qwen3-VL-4B-Instruct-MLX-6bit | quantized | MLX | 4.44B | M | apache-2.0 | lmstudio-community | 2 | 2.6M |
| lmstudio-community/Qwen3-VL-4B-Instruct-MLX-5bit | quantized | MLX | 4.44B | M | apache-2.0 | lmstudio-community | 1 | 2.6M |
| Comfy-Org/Qwen3-VL | finetune | — | — | — | apache-2.0 | Comfy-Org | 105 | 2.2M |
| Qwen/Qwen3-VL-4B-Instruct-FP8 | quantized | -FP8 | 4.83B | M | apache-2.0 | Qwen | 74 | 2M |
| unsloth/Qwen3-VL-4B-Instruct-GGUF | quantized | GGUF | — | — | apache-2.0 | unsloth | 78 | 1.9M |
| cyankiwi/Qwen3-VL-4B-Instruct-AWQ-4bit | quantized | AWQ | 4.83B | M | apache-2.0 | cyankiwi | 10 | 1.6M |
| unsloth/Qwen3-VL-4B-Instruct-unsloth-bnb-4bit | quantized | -bnb | 4.44B | M | apache-2.0 | unsloth | 12 | 1M |
| Qwen/Qwen3-VL-4B-Instruct-GGUF | quantized | GGUF | — | — | apache-2.0 | Qwen | 78 | 938.2k |
| TomoroAI/tomoro-colqwen3-embed-4b | finetune | — | 4.44B | M | apache-2.0 | TomoroAI | 36 | 784.4k |
| lmstudio-community/Qwen3-VL-4B-Instruct-GGUF | quantized | GGUF | — | — | apache-2.0 | lmstudio-community | 9 | 675.4k |
| unsloth/Qwen3-VL-4B-Instruct | finetune | — | 4.44B | M | apache-2.0 | unsloth | 11 | 398k |
| OpenSearch-AI/Ops-Colqwen3-4B | finetune | — | 4.44B | M | apache-2.0 | OpenSearch-AI | 11 | 330.6k |
| huihui-ai/Huihui-Qwen3-VL-4B-Instruct-abliterated | finetune | — | 4.44B | M | apache-2.0 | huihui-ai | 117 | 142.1k |
| DreamEternal/MinerU-Popo | finetune | — | 4.44B | M | mit | DreamEternal | 5 | 143.7k |
| StarVLA/Qwen3-VL-4B-Instruct-Action | finetune | — | 4.44B | M | — | StarVLA | 3 | 100.6k |
| Goekdeniz-Guelmez/Josiefied-Qwen3-VL-4B-Instruct-abliterated-beta-v1 | finetune | — | 4.44B | M | — | Goekdeniz-Guelmez | 5 | 88.9k |
| bartowski/Qwen_Qwen3-VL-4B-Instruct-GGUF | quantized | GGUF | — | — | — | bartowski | 6 | 45.4k |
| prithivMLmods/Qwen3-VL-4B-Instruct-abliterated-v1 | finetune | — | 4.44B | M | apache-2.0 | prithivMLmods | 27 | 34.1k |
| unsloth/Qwen3-VL-4B-Instruct-1M-GGUF | quantized | GGUF | — | — | apache-2.0 | unsloth | 6 | 33.3k |
| cyankiwi/Qwen3-VL-4B-Instruct-AWQ-8bit | quantized | AWQ | 2.22B | M | apache-2.0 | cyankiwi | 4 | 31.1k |
| mobilint/Qwen3-VL-4B-Instruct | quantized | — | 391M | S | apache-2.0 | mobilint | 0 | 31k |
| unsloth/Qwen3-VL-4B-Instruct-bnb-4bit | quantized | -bnb | 4.44B | M | apache-2.0 | unsloth | 3 | 30.3k |
| JEILDLWLRMA/Qwen3-VL-4B-Instruct-NVFP4 | quantized | NVFP4 | 3.24B | M | apache-2.0 | JEILDLWLRMA | 2 | 30.3k |
| DreamFast/Qwen3-VL-4b-Heretic-GGUF | quantized | GGUF | — | — | apache-2.0 | DreamFast | 17 | 29.5k |
| stepfun-ai/GELab-Zero-4B-preview | finetune | — | 4.44B | M | apache-2.0 | stepfun-ai | 153 | 20.4k |
| aisingapore/Qwen-SEA-LION-v4-4B-VL | finetune | — | 4.44B | M | mit | aisingapore | 6 | 19.3k |
| prithivMLmods/Qwen3-VL-4B-Instruct-Unredacted-MAX | finetune | — | 4.44B | M | apache-2.0 | prithivMLmods | 20 | 17.3k |
| Mungert/GELab-Zero-4B-preview-GGUF | quantized | GGUF | — | — | apache-2.0 | Mungert | 0 | 17.3k |
| robometer/Robometer-4B | finetune | — | 4.45B | M | apache-2.0 | robometer | 8 | 15.4k |
| tanhuajie2001/Robo-Dopamine-GRM-2.0-4B-Preview | finetune | — | 4.83B | M | apache-2.0 | tanhuajie2001 | 1 | 11k |
| furiosa-ai/Qwen3-VL-4B-Instruct | finetune | — | 4.44B | M | apache-2.0 | furiosa-ai | 1 | 10.7k |
| huihui-ai/Huihui-Qwen3-VL-4B-Instruct-abliterated-FP8 | quantized | -FP8 | 4.44B | M | apache-2.0 | huihui-ai | 14 | 9.4k |
| mradermacher/Qwen3-VL-4B-Instruct-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 1 | 9.7k |
| mradermacher/Qwen3-VL-4B-Instruct-i1-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 2 | 8.2k |
| teetone/RoboReward-4B | finetune | — | 570K | S | cc-by-4.0 | teetone | 9 | 7.8k |
| DreamFast/Qwen3-VL-4b-Heretic-ComfyUI | quantized | — | — | — | apache-2.0 | DreamFast | 163 | 0 |
| inclusionAI/ZwZ-4B | finetune | — | 4.83B | M | apache-2.0 | inclusionAI | 34 | 5.9k |
| SberRoboticsCenter/Qwen3-VL-4B-Instruct-action | finetune | — | 4.44B | M | — | SberRoboticsCenter | 0 | 6.9k |
| Date | Commit | SHA |
|---|---|---|
| 2025-10-15 | Adding `transformers` as the library. (#5) | ebb281e |
| 2025-10-15 | Fix: move inputs to model device in inference example to avoid "same device" error (#3) | da442ca |
| 2025-10-15 | Update README.md (#4) | 8c27028 |
| 2025-10-15 | Update README.md (#2) | a22a2ba |
| 2025-10-15 | Update README.md | 089b196 |
| 2025-10-14 | Update README.md | f2981ed |
| 2025-10-13 | Update README.md | 70c41a9 |
| 2025-10-11 | Upload folder using huggingface_hub | d51ae92 |
| 2025-10-11 | initial commit | 264f17f |
Hosted endpoints observed for Qwen3-VL-4B-Instruct, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| Featherless hf | Qwen/Qwen3-VL-4B-Instruct | — | — | — | — | — | — | — | live | 2026-10-25 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:33 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:00 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
