Up to date
llava-hf Apache 2.0 out 3 years ago
7.57B parameters, images and text in, text out, 32K tokens of context.
Verified 2026-10-25 Rank #546 in LLMs · score 75 method v1 Sources 3 Downloads 30d 256.4k · #229 of 619 tracked
Picked by rule from 2 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 1 pick above, plus these 1. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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: 4 quantized builds and 67 Spaces.
Top Spaces running it: merve/llava-next ♥ 151 · merve/vision_papers ♥ 118 · TencentARC/BrushEdit ♥ 69 · zamal/Multimodal-Chat-PDF ♥ 12
Weights only, an estimate: leave room for the KV cache. Method
About 4.5 GB at INT4, 9.1 GB at FP8, 18 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 (2024-02-20), amber triangles are 2 dated articles (each links to it), grey ticks are repo commits.
206 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 | LlavaNextForConditionalGeneration |
| Model type | llava_next |
| Dtype | float16 |
| Params by dtype | F16: 7,566,747,648 |
| Size band | M |
| MoE | no |
| Context length | 32,768 (config.json max_position_embeddings) |
| Languages | en |
| Papers | arXiv:2310.03744 |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · LlavaNextForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: not in registry · 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 llava-v1.6-mistral-7b-hf: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 18 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 9.1 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 4.5 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
finetune: 189quantized: 4adapter: 12merge: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| TIGER-Lab/VLM2Vec-LLaVa-Next | finetune | — | 7.57B | M | apache-2.0 | TIGER-Lab | 1 | 87.6k |
| unsloth/llava-v1.6-mistral-7b-hf-bnb-4bit | quantized | -bnb | 7.80B | M | apache-2.0 | unsloth | 9 | 61.6k |
| BAAI/BGE-VL-MLLM-S1 | finetune | — | 7.57B | M | mit | BAAI | 23 | 55.4k |
| BAAI/BGE-VL-v1.5-mmeb | finetune | — | 7.57B | M | mit | BAAI | 13 | 10.6k |
| unsloth/llava-v1.6-mistral-7b-hf | finetune | — | 7.57B | M | apache-2.0 | unsloth | 4 | 9.1k |
| BAAI/BGE-VL-MLLM-S2 | finetune | — | 7.57B | M | mit | BAAI | 17 | 6.5k |
| DeepGlint-AI/UniME-LLaVA-1.6-7B | finetune | — | 7.57B | M | mit | DeepGlint-AI | 5 | 2.8k |
| BAAI/BGE-VL-v1.5-zs | finetune | — | 7.57B | M | mit | BAAI | 9 | 1.9k |
| parasail-ai/VLM2Vec-LLaVa-Next-vllm | finetune | — | 7.57B | M | apache-2.0 | parasail-ai | 2 | 841 |
| OpenVINO/llava-v1.6-mistral-7b-hf-int8-ov | quantized | int8 | — | — | apache-2.0 | OpenVINO | 0 | 844 |
| Mantis-VL/llava_next_mistral_7b_4096 | finetune | — | 7.57B | M | — | Mantis-VL | 2 | 710 |
| thinhM5125024/llava-next-flickr30k-qlora-ckpt | adapter | — | — | — | apache-2.0 | thinhM5125024 | 1 | 526 |
| OpenVINO/llava-v1.6-mistral-7b-hf-int4-ov | quantized | int4 | — | — | apache-2.0 | OpenVINO | 2 | 387 |
| JUNJIE99/MMRet-MLLM-S1 | finetune | — | 7.57B | M | mit | JUNJIE99 | 0 | 330 |
| 2dameneko/llava-v1.6-mistral-7b-hf-nf4 | quantized | -nf4 | 7.80B | M | apache-2.0 | 2dameneko | 0 | 303 |
| Narrator5000/llavanext-finetuned-stackoverflow-vqa | adapter | — | — | — | apache-2.0 | Narrator5000 | 1 | 211 |
| UoM-CS-NeuroSymbolicAI/llava_next_mistral_math_10k | finetune | — | 603K | S | other | UoM-CS-NeuroSymbolicAI | 1 | 205 |
| aliencaocao/llava-1.6-mistral-7b-offensive-meme-singapore | finetune | — | 7.57B | M | mit | aliencaocao | 0 | 216 |
| JUNJIE99/MMRet-MLLM-S2 | finetune | — | 7.57B | M | mit | JUNJIE99 | 0 | 171 |
| tomaarsen/BGE-VL-MLLM-S1 | finetune | — | 7.57B | M | mit | tomaarsen | 0 | 170 |
| htlou/mm-interp-RLAIF-V-Dataset-llava-mistral | finetune | — | 7.57B | M | other | htlou | 0 | 162 |
| tomaarsen/BGE-VL-MLLM-S2 | finetune | — | 7.57B | M | mit | tomaarsen | 0 | 155 |
| htlou/mm-interp-AA_text_image_to_text | finetune | — | 7.57B | M | other | htlou | 0 | 121 |
| htlou/mm-interp-RLAIF-V-L0-q0_25_preference | finetune | — | 7.57B | M | other | htlou | 0 | 119 |
| htlou/mm-interp-RLAIF-V-Dataset | finetune | — | 7.57B | M | other | htlou | 0 | 116 |
| htlou/mm-interp-RLAIF-V_Coocur-q0_50 | finetune | — | 7.57B | M | other | htlou | 0 | 113 |
| htlou/mm-interp-RLAIF-V-Cosi-q0_75 | finetune | — | 7.57B | M | other | htlou | 0 | 107 |
| htlou/mm-interp-AA_preference_cocour_0_50 | finetune | — | 7.57B | M | other | htlou | 0 | 106 |
| htlou/mm-interp-AA_preference_cocour_new_step10_0_10 | finetune | — | 7.57B | M | other | htlou | 0 | 103 |
| htlou/mm-interp-AA_preference_l0_0_75 | finetune | — | 7.57B | M | other | htlou | 0 | 102 |
| htlou/mm-interp-AA_preference_random_0_80 | finetune | — | 7.57B | M | other | htlou | 0 | 101 |
| htlou/mm-interp-AA_preference_cocour_new_step10_0_60 | finetune | — | 7.57B | M | other | htlou | 0 | 100 |
| htlou/mm-interp-AA_preference_l0_new_step10_0_20 | finetune | — | 7.57B | M | other | htlou | 0 | 100 |
| htlou/mm-interp-AA_preference_cocour_new_step10_0_30 | finetune | — | 7.57B | M | other | htlou | 0 | 99 |
| htlou/mm-interp-AA_preference_cocour_new_step10_0_100 | finetune | — | 7.57B | M | other | htlou | 0 | 98 |
| htlou/mm-interp-AA_preference_l0_new_step10_0_70 | finetune | — | 7.57B | M | other | htlou | 0 | 98 |
| htlou/mm-interp-AA_preference_random_0_90 | finetune | — | 7.57B | M | other | htlou | 0 | 98 |
| htlou/mm-interp-AA_preference_cosi_new_step10_0_50 | finetune | — | 7.57B | M | other | htlou | 0 | 97 |
| htlou/mm-interp-AA_preference_l0_new_step10_0_60 | finetune | — | 7.57B | M | other | htlou | 0 | 97 |
| htlou/mm-interp-Align-Anything-Cosi-q0_25 | finetune | — | 7.57B | M | other | htlou | 0 | 95 |
| Date | Commit | SHA |
|---|---|---|
| 2025-12-22 | Fix dtype for the language model (#46) | 2424fdd |
| 2025-05-01 | Fixed a syntax error in code for "From transformers>=v4.48...." (#44) | 52320fb |
| 2025-02-10 | Fix typo in chat template (assistant EOS token) (#43) | 144bfb9 |
| 2025-02-06 | Fix typo in README (#42) | bb34097 |
| 2025-01-27 | Update for chat template | 0f72cba |
| 2025-01-08 | Update pipeline example | 5cfc077 |
| 2024-11-22 | add image as special token | ed45ac3 |
| 2024-11-19 | Create processor_config.json | e361ef3 |
| 2024-11-19 | checking endpoints demo example | 0cfb71e |
| 2024-10-21 | histiry -> history (#33) | 2f7f20b |
| 2024-09-19 | support system msg | 8a7baf5 |
| 2024-08-16 | update processor kwargs | 216670a |
| 2024-07-20 | fix escape char | 75e686c |
| 2024-07-19 | Update README.md | d9c4295 |
| 2024-07-10 | Create chat_template.json | 33884fe |
| 2024-06-28 | Delete processor_config.json | 05859c7 |
| 2024-06-28 | Upload processor | 3a3ab43 |
| 2024-06-18 | Delete processor_config.json | 213742d |
| 2024-06-18 | Upload processor | a39ec3e |
| 2024-05-09 | Update license to match upstream (Apache) (#15) | a1d5213 |
| 2024-03-21 | Update README.md | d7f9c48 |
| 2024-03-20 | Update README.md | c2e74c6 |
| 2024-03-20 | Update README.md | c509a63 |
| 2024-03-17 | Update README.md | dcc76c9 |
| 2024-03-17 | Update README.md | 9076059 |
| 2024-03-17 | Upload processor | 68f64ed |
| 2024-03-17 | Upload LlavaNextForConditionalGeneration | f71b5f1 |
| 2024-03-17 | Upload processor | e01056c |
| 2024-03-17 | Upload LlavaNextForConditionalGeneration | dfa9688 |
| 2024-03-17 | Upload processor | f5c511b |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency.
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 21:40 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:12 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.
