Up to date
llava-hf Llama 2 Community License out 3 years ago
7.06B parameters, images and text in, text out, 4K tokens of context.
Verified 2026-10-25 Rank #485 in LLMs · score 76.9 method v1 Sources 3 Downloads 30d 874.6k · #123 of 619 tracked
Changed this week: 1 new quantized build on the Hub (4 in all).2026-10-24
Picked by rule from 3 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Released in October 2023, LLaVA-1.5 accepts image and text input and can be trained on a single 8-A100 GPU node.
The 1 pick above, plus these 2. 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 100 Spaces.
Top Spaces running it: llava-hf/llava-4bit ♥ 60 · SeaLLMs/SeaLLM-Chat ♥ 38 · multimodalart/khala ♥ 31 · TIGER-Lab/Mantis ♥ 26
Weights only, an estimate: leave room for the KV cache. Method
About 4.2 GB at INT4, 8.5 GB at FP8, 17 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 (2023-12-05), amber triangles are 3 dated articles (each links to it), grey ticks are repo commits.
331 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 | LlavaForConditionalGeneration |
| Model type | llava |
| Dtype | float16 |
| Params by dtype | F16: 7,063,427,072 |
| Size band | M |
| MoE | no |
| Context length | 4,096 (config.json max_position_embeddings) |
| Languages | en |
| License | Llama 2 Community License (llama2) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · LlavaForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · LlavaForConditionalGeneration · 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-1.5-7b-hf: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 17 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 8.5 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 4.2 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 4finetune: 149adapter: 178
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| unsloth/llava-1.5-7b-hf-bnb-4bit | quantized | -bnb | 7.28B | M | llama2 | unsloth | 7 | 1.5M |
| unsloth/llava-1.5-7b-hf | finetune | — | 7.06B | M | llama2 | unsloth | 2 | 8.2k |
| Marwan02/llava-1.5-7b-hf-Q4_K_M-GGUF | quantized | GGUF | — | — | llama2 | Marwan02 | 1 | 3.4k |
| TIGER-Lab/Mantis-llava-7b | finetune | — | 7.06B | M | apache-2.0 | TIGER-Lab | 16 | 2.2k |
| HuggingFaceH4/sft-llava-1.5-7b-hf | adapter | — | — | — | llama2 | HuggingFaceH4 | 0 | 836 |
| YuchengShi/LLaVA-v1.5-7B-CUB-200 | finetune | — | 7.06B | M | — | YuchengShi | 4 | 508 |
| mohbattharani/LLaVA-Chef | finetune | — | 7.05B | M | llama3.1 | mohbattharani | 1 | 618 |
| Mantis-VL/llava_1.5_7b_v2_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 543 |
| astro189/working | adapter | — | — | — | llama2 | astro189 | 0 | 493 |
| shashikanth-a/llava-1.5-7b-hf-4bit | finetune | -4bit | 7.06B | M | llama2 | shashikanth-a | 0 | 487 |
| NouRed/Med-LLaVa-QLoRA | adapter | — | — | — | — | NouRed | 2 | 380 |
| Mantis-VL/mllava_llava_debug_nlvr2_v5_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 468 |
| georgeNakayama/AIpparel | finetune | — | — | — | mit | georgeNakayama | 9 | 0 |
| mohit311/LLaVA-data-json | adapter | — | — | — | llama2 | mohit311 | 0 | 439 |
| Salmamoori/llava-lastframe-finetuned | adapter | — | 7.36B | M | — | Salmamoori | 0 | 389 |
| colossatr0n/llava-1.5-7b-hf-Q8_0-GGUF | quantized | Q8_0 | — | — | llama2 | colossatr0n | 0 | 358 |
| Mantis-VL/mllava_v1_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 348 |
| edbeeching/vsft-llava-1.5-7b-hf | finetune | — | 7.06B | M | — | edbeeching | 0 | 298 |
| lfeng0722/llava-1.5-7b-hf-ft-mix-vsft | adapter | — | — | — | llama2 | lfeng0722 | 0 | 270 |
| Leonardo6/sft-llava-1.5-7b-hf | finetune | — | 7.06B | M | — | Leonardo6 | 1 | 217 |
| huggingface-mic/llava-mic | finetune | — | 4.34M | S | — | huggingface-mic | 0 | 257 |
| kkk5/CIM-LLaVA1.5-7B-SFT | finetune | — | 7.06B | M | llama2 | kkk5 | 0 | 247 |
| Mantis-VL/mllava_v2_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 231 |
| Mantis-VL/mllava_nlvr2_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 229 |
| edbeeching/vsft-llava-1.5-7b-hf8 | finetune | — | 7.06B | M | — | edbeeching | 0 | 228 |
| Cloudriver/MSD-LLaVA1.5-7B | finetune | — | — | — | apache-2.0 | Cloudriver | 0 | 228 |
| CrystalRaindropsFall/llava-l1-70pct | finetune | — | 7.06B | M | — | CrystalRaindropsFall | 1 | 167 |
| edbeeching/vsft-llava-1.5-7b-hf5 | finetune | — | 7.06B | M | — | edbeeching | 0 | 213 |
| Leonardo6/llava-sft-freeze | finetune | — | 599K | S | — | Leonardo6 | 0 | 198 |
| onlineshihab/fake-artwork-llava-lora | adapter | — | — | — | — | onlineshihab | 1 | 141 |
| metchee/persrv | adapter | — | — | — | — | metchee | 0 | 186 |
| Mantis-VL/mllava_debug_nlvr2_v2_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 183 |
| Mantis-VL/mllava_debug_nlvr2_v4_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 172 |
| PranavBP525/slava | adapter | — | 1.86B | S | — | PranavBP525 | 0 | 167 |
| Mantis-VL/mllava_debug_nlvr2_v5_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 163 |
| Leonardo6/llava-sft | finetune | — | 599K | S | — | Leonardo6 | 0 | 152 |
| Mantis-VL/mllava_llava_debug_nlvr2_v6_4096 | finetune | — | 7.06B | M | — | Mantis-VL | 0 | 150 |
| 2toINF/IVM | finetune | — | — | — | apache-2.0 | 2toINF | 3 | 0 |
| simaai/llava-1.5-7b-hf-a16w4 | finetune | — | — | — | mit | simaai | 0 | 150 |
| Salmamoori/llava-bddx-finetuned | adapter | — | 7.36B | M | — | Salmamoori | 2 | 49 |
| Date | Commit | SHA |
|---|---|---|
| 2025-06-06 | Create chat_template.jinja | b234b80 |
| 2025-01-27 | typo ... | 6ceb2ed |
| 2025-01-27 | Add example for vectorized chat template | bdb8c4a |
| 2025-01-27 | Enable inference | fc7f5bf |
| 2025-01-13 | histiry -> history (#45) | e2214c2 |
| 2025-01-08 | Update pipeline example | f717a66 |
| 2024-11-22 | add image as special token | 8c85e9a |
| 2024-11-18 | Upload processor | ca4bea3 |
| 2024-09-14 | Update README.md | a272c74 |
| 2024-09-14 | Update README.md | 79a2b54 |
| 2024-09-14 | Update README.md (#32) | 8bb1631 |
| 2024-09-03 | Typo (#30) | 95f2cf3 |
| 2024-08-16 | update processor kwargs | fa3dd28 |
| 2024-08-13 | update pipeline tag | 37a8553 |
| 2024-08-13 | Update chat_template.json to incorporate `generation` tag (#29) | 73d3444 |
| 2024-07-26 | Fix system message | 9e7ca49 |
| 2024-07-19 | Add chat template examples | 2f2b2b1 |
| 2024-07-10 | Create chat_template.json | 17b3a43 |
| 2024-06-28 | Delete processor_config.json | 1695216 |
| 2024-06-28 | Upload processor | 42a0f23 |
| 2024-06-18 | Delete processor_config.json | 9bda95c |
| 2024-06-18 | Upload processor | bb939be |
| 2024-01-30 | Update text config to Vicuna (#15) | 05ae243 |
| 2023-12-26 | Update README.md | 6b71355 |
| 2023-12-15 | Set `tie_word_embeddings` to False (#7) | 3685f18 |
| 2023-12-09 | Update README.md | 0db0c8c |
| 2023-12-09 | Update README.md | d5e0832 |
| 2023-12-09 | Update README.md | 1acd332 |
| 2023-12-07 | Upload processor | cb14093 |
| 2023-12-07 | Update tokenizer_config.json | 03ffc82 |
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.
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 15:11 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.
