Up to date
3.82B parameters, text in, text out, 128K tokens of context.
Verified 2026-10-25 Rank #168 in LLMs · score 91.3 method v1 Sources 3 Downloads 30d 375.8k · #188 of 619 tracked
Changed this week: 5 new quantized builds on the Hub (201 in all).2026-10-24
Picked by rule from 10 tracked pieces: the lab's own post, the most shared, a field report, the latest.
BestLLMfor operates as a meta-review of AI tools rather than a single model assessment, prioritizing free local LLMs unless online services offer unique features or exceed local hardware capabilities.
The 1 pick above, plus these 9. 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: 201 quantized builds and 100 Spaces.
Top Spaces running it: pliny-the-prompter/obliteratus ♥ 621 · cvachet/pdf-chatbot ♥ 382 · baconnier/prompt-plus-plus ♥ 246 · AiActivity/AI-Assistant ♥ 21
Weights only, an estimate: leave room for the KV cache. Method
About 2.3 GB at INT4, 4.6 GB at FP8, 9.2 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 (2024-08-16), amber triangles are 6 dated articles (each links to it), grey ticks are repo commits.
1,276 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.
Phi-3.5-mini-instruct is a 3.82B parameter text-generation model from the Phi-3 model family.[1][2] Released under an MIT license, this multilingual model supports a 128K token context length.[3][4]
Generated 2026-10-07 15:04 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
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 | text-generation |
|---|---|
| Library | transformers |
| Architectures | Phi3ForCausalLM |
| Model type | phi3 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 3,821,079,552 |
| Size band | M |
| MoE | no |
| Context length | 131,072 (config.json max_position_embeddings) |
| Languages | multilingual |
| Papers | arXiv:2404.14219, arXiv:2407.13833, arXiv:2403.06412 |
| License | MIT (mit) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Phi3ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Phi3ForCausalLM · 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 Phi-3.5-mini-instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 9.2 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 4.6 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 2.3 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 201finetune: 340adapter: 716merge: 19
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| MaziyarPanahi/Phi-3.5-mini-instruct-GGUF | quantized | GGUF | — | — | — | MaziyarPanahi | 34 | 12.4M |
| numind/NuExtract-1.5 | finetune | — | 3.82B | M | mit | numind | 250 | 1.4M |
| bartowski/Phi-3.5-mini-instruct-GGUF | quantized | GGUF | — | — | mit | bartowski | 94 | 896.7k |
| unsloth/Phi-3.5-mini-instruct-bnb-4bit | quantized | -bnb | 3.82B | M | mit | unsloth | 14 | 619.2k |
| flowaicom/Flow-Judge-v0.1-AWQ | quantized | AWQ | 3.82B | M | apache-2.0 | flowaicom | 6 | 437k |
| zstanjj/HTML-Pruner-Phi-3.8B | finetune | — | 3.82B | M | apache-2.0 | zstanjj | 19 | 235.4k |
| unsloth/Phi-3.5-mini-instruct | finetune | — | 3.82B | M | mit | unsloth | 53 | 217.3k |
| anakin87/Phi-3.5-mini-ITA | finetune | — | 3.82B | M | mit | anakin87 | 13 | 145.6k |
| mlc-ai/Phi-3.5-mini-instruct-q4f16_1-MLC | quantized | — | — | — | — | mlc-ai | 5 | 131.7k |
| ReDiX/Artemide-3.5 | finetune | — | 3.82B | M | mit | ReDiX | 3 | 127.5k |
| mlc-ai/Phi-3.5-mini-instruct-q4f16_0-MLC | quantized | — | — | — | — | mlc-ai | 0 | 88.9k |
| flowaicom/Flow-Judge-v0.1 | finetune | — | 3.82B | M | apache-2.0 | flowaicom | 73 | 79.8k |
| microsoft/MediPhi-Instruct | finetune | — | 3.82B | M | mit | microsoft | 75 | 51.8k |
| microsoft/MediPhi-Clinical | finetune | — | 3.82B | M | mit | microsoft | 23 | 34k |
| PatronusAI/glider | finetune | — | 3.82B | M | cc-by-nc-4.0 | PatronusAI | 47 | 25.8k |
| microsoft/MediPhi | finetune | — | 3.82B | M | mit | microsoft | 29 | 20.6k |
| legraphista/Phi-3.5-mini-instruct-IMat-GGUF | quantized | GGUF | — | — | mit | legraphista | 2 | 21.7k |
| onnx-community/Phi-3.5-mini-instruct-onnx-web | quantized | — | — | — | mit | onnx-community | 15 | 20.9k |
| amd/Phi-3.5-mini-instruct-awq-g128-int4-asym-bf16-onnx-ryzen-strix | quantized | awq | — | — | mit | amd | 3 | 21.4k |
| OpenVINO/Phi-3.5-mini-instruct-int4-ov | quantized | int4 | — | — | mit | OpenVINO | 10 | 19k |
| AcerTest359/Phi-3.5-mini-instruct-OV | finetune | — | — | — | mit | AcerTest359 | 0 | 13.8k |
| grounded-ai/phi3.5-hallucination-judge | adapter | — | — | — | mit | grounded-ai | 9 | 13.3k |
| cowoho/Phi-3.5-mini-instruct-Q4_K_M-GGUF | quantized | GGUF | — | — | mit | cowoho | 1 | 13.5k |
| flockgo/task-5-microsoft-Phi-3-mini-4k-instruct | adapter | — | — | — | — | flockgo | 0 | 13.4k |
| second-state/Phi-3.5-mini-instruct-GGUF | quantized | GGUF | — | — | mit | second-state | 0 | 12k |
| Mungert/MediPhi-Instruct-GGUF | quantized | GGUF | — | — | mit | Mungert | 0 | 12k |
| latiao1999/task-6-microsoft-Phi-3-mini-4k-instruct | adapter | — | — | — | — | latiao1999 | 0 | 9.2k |
| srrthk/task-13-microsoft-Phi-3.5-mini-instruct | adapter | — | — | — | — | srrthk | 0 | 8k |
| OpenVINO/Phi-3.5-mini-instruct-int8-ov | quantized | int8 | — | — | mit | OpenVINO | 0 | 7.6k |
| QuantFactory/Phi-3.5-mini-ITA-GGUF | quantized | GGUF | — | — | mit | QuantFactory | 4 | 7.4k |
| lfhe/FLock-Arena-Task-9-Argus | adapter | — | — | — | — | lfhe | 0 | 7.5k |
| QuantFactory/calme-2.1-phi3.5-4b-GGUF | quantized | GGUF | — | — | mit | QuantFactory | 2 | 7.3k |
| microsoft/MediPhi-PubMed | finetune | — | 3.82B | M | mit | microsoft | 19 | 5.8k |
| tensorblock/Phi-3.5-mini-instruct-GGUF | quantized | GGUF | — | — | mit | tensorblock | 1 | 6.7k |
| QuantFactory/phi3.5-gutenberg-4B-GGUF | quantized | GGUF | — | — | mit | QuantFactory | 4 | 6k |
| AcerTest359/Phi-3.5-mini-instruct-GGUF | quantized | GGUF | — | — | mit | AcerTest359 | 0 | 6.1k |
| mradermacher/Phi-3.5-mini-instruct-i1-GGUF | quantized | GGUF | — | — | mit | mradermacher | 0 | 5.8k |
| lfhe/FLock-Arena-Task-15-Carbonia | adapter | — | — | — | — | lfhe | 0 | 5.7k |
| OpenVINO/Phi-3.5-mini-instruct-int4-cw-ov | quantized | int4 | — | — | mit | OpenVINO | 3 | 5.4k |
| MaziyarPanahi/calme-2.1-phi3.5-4b | finetune | — | 3.82B | M | mit | MaziyarPanahi | 4 | 5.3k |
| Date | Commit | SHA |
|---|---|---|
| 2025-12-10 | Update README.md (#43) | 2fe1924 |
| 2025-12-10 | Upload data_summary_card.md (#42) | a74e37c |
| 2025-03-02 | Update readme | 3145e03 |
| 2024-09-18 | Add Onnx model url | af0dfb8 |
| 2024-08-29 | Update README.md | ccf028f |
| 2024-08-29 | Update README.md. Added Korean benchmarks. (#18) | bdf5de1 |
| 2024-08-22 | Update README.md | cd6881a |
| 2024-08-22 | Update README.md | 8ddd775 |
| 2024-08-20 | Update README.md | 6496300 |
| 2024-08-20 | Update README.md | 5adb202 |
| 2024-08-20 | upload model card | d509c56 |
| 2024-08-17 | upload weights | 02b3eb6 |
| 2024-08-17 | Upload initial files | 7a1ac0d |
| 2024-08-16 | initial commit | f27a2f3 |
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 21:36 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:04 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.
