Up to date
google health-ai-developer-foundations out 17 months ago
4.30B parameters, images and text in, text out.
Verified 2026-10-25 Rank #146 in LLMs · score 93.5 method v1 Sources 3 Downloads 30d 430.8k · #177 of 619 tracked
Picked by rule from 10 tracked pieces: the lab's own post, the most shared, a field report, the latest.
MedGemma 4B IT is a multimodal Google language model released in May 2025 designed for medical text and image comprehension.
Google releases MedGemma 1.5 4B, an updated open medical model supporting high-dimensional imaging like CT and MRI, alongside a new speech-to-text model called MedASR.
The 2 picks above, plus these 8. 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: 56 quantized builds and 100 Spaces.
Top Spaces running it: google/rad_explain ♥ 248 · warshanks/medgemma-4b-it ♥ 40 · rishiraj/radiology ♥ 20 · khalednabawi11/Medical-Scan-Gemma ♥ 14
Weights only, an estimate: leave room for the KV cache. Method
About 2.6 GB at INT4, 5.2 GB at FP8, 10 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 (2025-05-19), amber triangles are 5 dated articles (each links to it).
812 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 | Gemma3ForConditionalGeneration |
| Model type | gemma3 |
| Params by dtype | BF16: 4,300,079,472 |
| Size band | M |
| MoE | no |
| Gated | auto |
| Papers | arXiv:2303.15343, arXiv:2507.05201, arXiv:2405.03162, arXiv:2106.14463, arXiv:2412.03555, arXiv:2501.19393, arXiv:2009.13081, arXiv:2102.09542, arXiv:2411.15640, arXiv:2404.05590, arXiv:2501.18362 |
| License | health-ai-developer-foundations (other) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Gemma3ForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Gemma3ForConditionalGeneration · 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 medgemma-4b-it: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 10 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 5.2 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 2.6 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Base model: google/medgemma-4b-pt
quantized: 56finetune: 629merge: 3adapter: 124
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| unsloth/medgemma-4b-it-GGUF | quantized | GGUF | — | — | other | unsloth | 75 | 289.4k |
| unsloth/medgemma-4b-it-unsloth-bnb-4bit | quantized | -bnb | 4.39B | M | other | unsloth | 1 | 87.7k |
| bartowski/google_medgemma-4b-it-GGUF | quantized | GGUF | — | — | other | bartowski | 6 | 51.1k |
| unsloth/medgemma-4b-it | finetune | — | 4.97B | M | other | unsloth | 7 | 34.1k |
| unsloth/medgemma-4b-it-bnb-4bit | quantized | -bnb | 4.41B | M | other | unsloth | 5 | 20.5k |
| DevQuasar/google.medgemma-4b-it-GGUF | quantized | GGUF | — | — | — | DevQuasar | 2 | 15.2k |
| lmstudio-community/medgemma-4b-it-MLX-4bit | finetune | MLX | 4.97B | M | other | lmstudio-community | 2 | 15k |
| lmstudio-community/medgemma-4b-it-GGUF | quantized | GGUF | — | — | other | lmstudio-community | 6 | 14k |
| mradermacher/medgemma-4b-it-i1-GGUF | quantized | GGUF | — | — | other | mradermacher | 1 | 13.7k |
| second-state/medgemma-4b-it-GGUF | quantized | GGUF | — | — | other | second-state | 1 | 13.2k |
| rajpurkarlab/medgemma-4b-it-crimson | finetune | — | 3.88B | M | other | rajpurkarlab | 3 | 8k |
| SaltySander/ANTONI-Alpha | finetune | — | — | — | other | SaltySander | 8 | 6.9k |
| axel-darmouni/medgemma-4b-it-sft-brain-regions | finetune | — | — | — | — | axel-darmouni | 0 | 7.1k |
| mradermacher/medgemma-4b-it-GGUF | quantized | GGUF | — | — | other | mradermacher | 0 | 6.9k |
| SandLogicTechnologies/MedGemma-4B-IT-GGUF | quantized | GGUF | — | — | gemma | SandLogicTechnologies | 5 | 5.4k |
| Sadou/medgemma-4b-it-french-medical-assistant | finetune | — | — | — | apache-2.0 | Sadou | 1 | 5.6k |
| ekacare/parrotlet-a-en-5b | finetune | — | — | — | mit | ekacare | 4 | 5k |
| Aakash010/MedGemma_FineTuned | quantized | GGUF | 4.30B | M | apache-2.0 | Aakash010 | 3 | 4.9k |
| XythicK/medgemma-4b-it-GGUF | quantized | GGUF | — | — | — | XythicK | 1 | 4.8k |
| Gunulhona/Gemma-3-4B | merge | — | 4.30B | M | — | Gunulhona | 1 | 4k |
| matrixportalx/Gemma-3-4B-GGUF | merge | GGUF | — | — | — | matrixportalx | 1 | 3.4k |
| lastmass/MedGemma-GRPO | finetune | — | 4.97B | M | apache-2.0 | lastmass | 2 | 3.1k |
| gaianet/medgemma-4b-it-GGUF | quantized | GGUF | — | — | other | gaianet | 0 | 3.2k |
| Tavorish/medgemma-4b-it-GGUF | quantized | GGUF | — | — | other | Tavorish | 0 | 3.2k |
| sciedoc/neurojenml-inference | finetune | — | 4.97B | M | — | sciedoc | 0 | 3k |
| jwest33/medgemma-4b-it-null-space-abliterated-GGUF | quantized | GGUF | — | — | gemma | jwest33 | 3 | 2.8k |
| ekacare/parrotlet-a-2.5-pro | finetune | — | — | — | other | ekacare | 4 | 2k |
| edziocodes/medgemma-breast-cancer | finetune | — | 4.97B | M | mit | edziocodes | 7 | 1.2k |
| ClinicDx1/ClinicDx | adapter | GGUF | 4.30B | M | gemma | ClinicDx1 | 3 | 1.3k |
| hungqbui/medgemma-4b-it-Q4_K_M-GGUF | quantized | GGUF | — | — | other | hungqbui | 0 | 1.4k |
| drwlf/medgemma-4b-it-abliterated | adapter | — | 4.97B | M | apache-2.0 | drwlf | 6 | 969 |
| geeek/medgemma-4b-it-ONNX | quantized | — | — | — | other | geeek | 1 | 1.2k |
| FastFlowLM/medgemma-4b-it-NPU2 | finetune | — | — | — | other | FastFlowLM | 0 | 1.1k |
| NikolayKozloff/medgemma-4b-it-Q8_0-GGUF | quantized | Q8_0 | — | — | other | NikolayKozloff | 3 | 860 |
| sciedoc/neurojenml | adapter | — | — | — | — | sciedoc | 0 | 966 |
| Balthulal/MedGemma_FineTuned | quantized | GGUF | 4.30B | M | apache-2.0 | Balthulal | 0 | 918 |
| pamessina/medgemma-4b-it-cure | adapter | — | — | — | other | pamessina | 0 | 882 |
| wangyichen25/medgemma-4b-it-sft-full-hyperkvasir | finetune | — | 4.30B | M | — | wangyichen25 | 0 | 879 |
| kingabzpro/medgemma-brain-cancer | finetune | — | — | — | apache-2.0 | kingabzpro | 16 | 0 |
| ahmet-harun/turk-rad-llm-lumbar-mini-v2 | adapter | — | — | — | other | ahmet-harun | 0 | 785 |
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 21:34 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:01 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.
