The story so far
Weights, every dated article, and repo commits on one rail.
15 repo commits since the weights landed
Teal is the weights (2025-04-10), grey ticks are repo commits.
InternVL3-8B
OpenGVLab Apache 2.0 (qwen) out 18 months ago
7.94B parameters, images and text in, text out.
Verified 2026-10-25 Rank #696 in LLMs · score 72.1 method v1 Sources 3 Downloads 30d 80.7k · #338 of 619 tracked
Do people use it?80.7kdownloads, 30 days#338 of 619 we track · below the median+0 likes, downloads −19% this week112 likesHugging Face2026-10-24
Where it's used
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.
No API provider serves it, so its downloads are people running it themselves: 5 quantized builds and 3 Spaces.
- As a service0no API provider listed
- On own hardware5quantized builds
- Built on5035 fine-tunes · 8 adapters · 7 merges
- In apps3Hugging Face Spaces
Top Spaces running it: TIGER-Lab/MEGA-Bench ♥ 51 · developer0hye/InternVL3-8B ♥ 6
Will it fit?
Weights only, an estimate: leave room for the KV cache. Method
Fits a 16 GB card at INT4
About 4.8 GB at INT4, 9.5 GB at FP8, 19 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
What people built on it
55 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
55 derivatives on the Hub, most of them fine-tunes
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Reference
Everything else, folded: the full tables, every source.
Use-case evidence
6 use cases
What it is good at 6 use cases
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.
- Understand images and videostructuraldecision-model estimate p=1.00tag: image-text-to-textpipeline: image-text-to-text+1 more
- Chat and follow instructionsstructuraldecision-model estimate p=0.99tag: conversationalchat template present
- Run on a laptop or phonestructuraldecision-model estimate p=0.71fits: single-gpu
- Embed text for searchstructuraltag: feature-extraction
- Work across languagesstructuraltag: multilingual
- Call tools and APIsstructuraldecision-model estimate p=0.87
Facts
19 fields
Listed in LLMs
Facts
- InternVL3-8B has 7.94B parameters. (safetensors metadata, observed 2026-10-24)
- License: Apache 2.0 (qwen) (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
- Created 2025-04-10, last modified 2025-09-11. (HF API, observed 2026-10-24)
| Pipeline tag | image-text-to-text |
|---|---|
| Library | transformers |
| Architectures | InternVLChatModel |
| Model type | internvl_chat |
| Dtype | bfloat16 |
| Params by dtype | BF16: 7,944,373,760 |
| Size band | M |
| MoE | no |
| Languages | multilingual |
| Papers | arXiv:2312.14238, arXiv:2404.16821, arXiv:2412.05271, arXiv:2411.10442, arXiv:2504.10479, arXiv:2412.09616 |
| License | Apache 2.0 (qwen) (apache-2.0) |
| Hub snapshot | 2026-10-25 |
Servability
vLLM · SGLang
Servability
vLLM: supported · InternVLChatModel · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · InternVLChatModel · 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.
Hardware fit
3 estimates
Hardware fit ESTIMATE
Weights-only memory for InternVL3-8B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 19 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 9.5 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 4.8 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
All variants
40 variants
Lineage
Base model: OpenGVLab/InternVL3-8B-Instruct
quantized: 5finetune: 35merge: 7adapter: 8
Top variants (40 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| OpenGVLab/InternVL3-8B-AWQ | quantized | AWQ | — | — | other | OpenGVLab | 8 | 728.3k |
| Thunderbolt215215/UniPercept | finetune | — | 7.94B | M | apache-2.0 | Thunderbolt215215 | 18 | 73.8k |
| unsloth/InternVL3-8B-GGUF | finetune | GGUF | — | — | apache-2.0 | unsloth | 6 | 34.5k |
| sensenova/SenseNova-SI-1.1-InternVL3-8B | finetune | — | 7.94B | M | apache-2.0 | sensenova | 16 | 23.3k |
| ByteDance/Sa2VA-Qwen3-VL-4B | merge | — | 5.06B | M | apache-2.0 | ByteDance | 20 | 22.4k |
| unsloth/InternVL3-8B | finetune | — | 7.94B | M | apache-2.0 | unsloth | 2 | 13.1k |
| mradermacher/InternVL3-8B-i1-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 0 | 12.2k |
| sensenova/SenseNova-SI-1.2-InternVL3-8B | finetune | — | 7.94B | M | apache-2.0 | sensenova | 11 | 10.7k |
| sensenova/SenseNova-SI-1.3-InternVL3-8B | finetune | — | 7.94B | M | apache-2.0 | sensenova | 11 | 10k |
| mradermacher/InternVL3-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 2 | 7.5k |
| InternSVG/InternSVG-8B | finetune | — | 7.95B | M | apache-2.0 | InternSVG | 4 | 6k |
| sensenova/SenseNova-SI-1.4-InternVL3-8B | finetune | — | 7.94B | M | apache-2.0 | sensenova | 4 | 5.7k |
| Thunderbolt215215/ArtiMuse | finetune | — | 7.94B | M | apache-2.0 | Thunderbolt215215 | 13 | 4.3k |
| ByteDance/Sa2VA-InternVL3-2B | merge | — | 2.32B | M | apache-2.0 | ByteDance | 2 | 3.5k |
| ByteDance/Sa2VA-Qwen3-VL-2B | merge | — | 2.67B | M | apache-2.0 | ByteDance | 18 | 2.1k |
| brandonbeiler/InternVL3-8B-FP8-Dynamic | quantized | -FP8 | 7.95B | M | mit | brandonbeiler | 2 | 2.9k |
| ByteDance/Sa2VA-Qwen2_5-VL-3B | merge | — | 4.29B | M | apache-2.0 | ByteDance | 3 | 2.9k |
| ByteDance/Sa2VA-Qwen2_5-VL-7B | merge | — | 8.53B | M | apache-2.0 | ByteDance | 5 | 2k |
| ByteDance/Sa2VA-InternVL3-14B | merge | — | 15.4B | L | apache-2.0 | ByteDance | 10 | 1.6k |
| ByteDance/Sa2VA-InternVL3-8B | merge | — | 8.18B | M | apache-2.0 | ByteDance | 6 | 1.5k |
| manglu3935/Chiron-o1-8B | finetune | — | 7.94B | M | mit | manglu3935 | 3 | 1.2k |
| OpenMOSS-Team/Game-RL-InternVL3-8B | finetune | — | 7.94B | M | apache-2.0 | OpenMOSS-Team | 2 | 352 |
| sensenova/SenseNova-SI-1.1-InternVL3-8B-800K | finetune | — | 7.94B | M | apache-2.0 | sensenova | 4 | 172 |
| widegather/unipercept-mirror | finetune | — | 7.94B | M | apache-2.0 | widegather | 0 | 349 |
| paulpacaud/guardian-thinking | finetune | — | 7.94B | M | apache-2.0 | paulpacaud | 0 | 271 |
| Thunderbolt215215/ArtiMuse_AVA | finetune | — | 7.94B | M | apache-2.0 | Thunderbolt215215 | 0 | 221 |
| paulpacaud/guardian-vanilla | finetune | — | 7.98B | M | apache-2.0 | paulpacaud | 0 | 212 |
| Thunderbolt215215/ArtiMuse_TAD66K | finetune | — | 7.94B | M | apache-2.0 | Thunderbolt215215 | 1 | 122 |
| fernandoruiz/InternVL3-8B-Q4_0-GGUF | finetune | Q4_0 | — | — | apache-2.0 | fernandoruiz | 0 | 168 |
| owl10/ReCogDrive | finetune | — | — | — | apache-2.0 | owl10 | 3 | 0 |
| MCG-NJU/LongVPO-Stage2-InternVL3-8B | finetune | — | 7.94B | M | mit | MCG-NJU | 0 | 126 |
| gw49/EMIT-8B | finetune | — | 7.94B | M | — | gw49 | 0 | 125 |
| MCG-NJU/LongVPO-Stage1-InternVL3-8B | finetune | — | 7.94B | M | mit | MCG-NJU | 0 | 122 |
| atahmih/InternVL3-8B-Q4_K_M-GGUF | finetune | GGUF | — | — | apache-2.0 | atahmih | 0 | 121 |
| Thunderbolt215215/ArtiMuse_PARA | finetune | — | 7.94B | M | apache-2.0 | Thunderbolt215215 | 0 | 111 |
| amd/InternVL3-8B-da8w8-torchao-v0.17.0 | quantized | — | — | — | mit | amd | 0 | 110 |
| whalezzz/MM-RAIT-InternVL_3 | finetune | — | 7.94B | M | mit | whalezzz | 0 | 102 |
| sarahyo941/UnifiedReward-Think-rejection_sampling_whole | finetune | — | 7.93B | M | apache-2.0 | sarahyo941 | 0 | 99 |
| muhammedsaidckr/InternVL3-8B-Q4_K_M-GGUF | finetune | GGUF | — | — | apache-2.0 | muhammedsaidckr | 1 | 43 |
| bangskitchen/EUEA_InternVL3-8B_ALFRED_SFT | finetune | — | 7.94B | M | mit | bangskitchen | 0 | 93 |
Repository changelog
15 commits
Repository changelog 15 commits
| Date | Commit | SHA |
|---|---|---|
| 2025-09-11 | Upload tokenizer_config.json with huggingface_hub | 853e3a7 |
| 2025-09-11 | Upload tokenizer_config.json with huggingface_hub | 9e7a9d5 |
| 2025-05-29 | Update README.md | 24dc81a |
| 2025-04-25 | Update README.md | 26dcade |
| 2025-04-17 | Upload README.md with huggingface_hub | 6f72873 |
| 2025-04-16 | Upload modeling_intern_vit.py with huggingface_hub | 5f6ae91 |
| 2025-04-15 | Upload modeling_internvl_chat.py with huggingface_hub | 99c6315 |
| 2025-04-15 | Upload README.md with huggingface_hub | 09ea619 |
| 2025-04-14 | Upload README.md with huggingface_hub | 291ac35 |
| 2025-04-12 | Upload README.md with huggingface_hub | c2d2995 |
| 2025-04-12 | Upload README.md with huggingface_hub | 04b6ad9 |
| 2025-04-12 | Upload README.md with huggingface_hub | 89ef9a8 |
| 2025-04-12 | Upload README.md with huggingface_hub | d10a145 |
| 2025-04-10 | Upload folder using huggingface_hub | dab7194 |
| 2025-04-10 | initial commit | 0a4db75 |
Papers
6 papers
Papers 6
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).
- V2PE: Improving Multimodal Long-Context Capability of Vision-Language Models with Variable Visual Position Encoding cited in card peer-reviewed · measuredJunqi Ge, Ziyi Chen, Jintao Lin +4 · 2024-12-12 · IEEE/CVF International Conference on Computer Vision (ICCV) · PDF · HF paper page (1 upvotes) · evidence
Similar models
3 models
Similar models
- Qwen3-VL-8B-Instruct · 8.77B same size band (M), same task (image-text-to-text), more providers, permissive license
- Qwen2.5-VL-7B-Instruct · 8.29B same size band (M), same task (image-text-to-text), more providers, permissive license
- Qwen3.5-9B · 9.65B same task (image-text-to-text), more providers, permissive license
Sources
3 sources
Sources
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:42 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:14 UTC · Last verified 2026-10-25
Follow InternVL3-8B
Watch this model
- /feeds/models/OpenGVLab--InternVL3-8B.xml(RSS 2.0) · every change to InternVL3-8B
- /feeds/families/OpenGVLab--InternVL3.xml(RSS 2.0) · every change across the InternVL3 family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.