Up to date
LiquidAI lfm1.0 out 20 days ago
3.12B parameters, images and text in, text out.
Verified 2026-10-25 Rank #474 in LLMs · score 77.3 method v1 Sources 3 Downloads 30d 26.5k · #435 of 619 tracked
Changed this week: 1 new quantized build on the Hub (21 in all).2026-10-24
Picked by rule from 20 tracked pieces: the lab's own post, the most shared, a field report, the latest.
ResolveHQ is a Cloudflare-native helpdesk where Cloudflare D1 holds tickets and customers, Cloudflare R2 holds attachments, and Cloudflare Queues carry inbound and outbound mail jobs.
The 1 pick above, plus these 19. 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: 21 quantized builds.
Weights only, an estimate: leave room for the KV cache. Method
About 1.9 GB at INT4, 3.7 GB at FP8, 7.5 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 (2026-10-05), amber triangles are 10 dated articles (each links to it).
37 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 | Lfm2VlForConditionalGeneration |
| Model type | lfm2_vl |
| Params by dtype | BF16: 3,123,483,888 |
| Size band | M |
| MoE | no |
| Papers | arXiv:2511.23404 |
| License | lfm1.0 (other) |
| Hub snapshot | 2026-10-25 |
Weights-only memory for d1-3B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 7.5 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 3.7 GB · Edge (<4 GB) |
| INT4 (0.5 bytes/param) | 1.9 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Base model: LiquidAI/LFM2.5-VL-3B
quantized: 21finetune: 13adapter: 3
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| LiquidAI/d1-3B-GGUF | quantized | GGUF | — | — | other | LiquidAI | 63 | 37.4k |
| NANI-Nithin/d1-3B-GGUF | quantized | GGUF | — | — | — | NANI-Nithin | 1 | 5k |
| AtomicChat/d1-3B-GGUF | quantized | GGUF | — | — | other | AtomicChat | 13 | 2.3k |
| webmp3/Sakura-d1-3B-HighQuality-GGUF | quantized | GGUF | — | — | other | webmp3 | 2 | 1.6k |
| LiquidAI/d1-3B-w8a8 | quantized | -w8a8 | 3.13B | M | other | LiquidAI | 10 | 721 |
| LiquidAI/d1-3B-MLX-fp16 | finetune | MLX | 3.12B | M | other | LiquidAI | 5 | 928 |
| prithivMLmods/d1-3B-GGUF | quantized | GGUF | — | — | other | prithivMLmods | 1 | 811 |
| LiquidAI/d1-3B-MLX-8bit | quantized | MLX | 3.12B | M | other | LiquidAI | 3 | 705 |
| LiquidAI/d1-3B-MLX-4bit | quantized | MLX | 3.12B | M | other | LiquidAI | 3 | 617 |
| TechnoBaptist/d1-3B-GGUF | quantized | GGUF | — | — | other | TechnoBaptist | 2 | 602 |
| DJLougen/d1-3B-MLX-8bit | quantized | MLX | 3.12B | M | other | DJLougen | 3 | 431 |
| onnx-community/d1-3B-ONNX | quantized | — | — | — | other | onnx-community | 3 | 197 |
| litert-community/d1-3B-LiteRT | finetune | — | — | — | other | litert-community | 1 | 249 |
| OceanLabs/Coral-2-3B | finetune | — | 3.12B | M | other | OceanLabs | 3 | 130 |
| OceanLabs/Coral-2-3B-fast | finetune | — | 3.12B | M | apache-2.0 | OceanLabs | 1 | 143 |
| yannelli/d1-3b-code-review-decisions | finetune | — | 3.12B | M | other | yannelli | 2 | 58 |
| mlboydaisuke/d1-3B-CoreAI | quantized | — | — | — | other | mlboydaisuke | 2 | 30 |
| WasamiKirua/liquid-d1-3b-memory-gate-GGUF | quantized | GGUF | — | — | other | WasamiKirua | 1 | 66 |
| wegotomars/d1-3B-CoreAI-ANE | finetune | — | — | — | other | wegotomars | 1 | 44 |
| RepublicOfKorokke/d1-3B-oQ4e-fp16 | quantized | -fp16 | 3.12B | M | — | RepublicOfKorokke | 0 | 78 |
| heath0xFF/touch-grass-v0 | finetune | — | 3.12B | M | other | heath0xFF | 1 | 24 |
| RepublicOfKorokke/d1-3B-oQ4e | quantized | — | 3.12B | M | — | RepublicOfKorokke | 0 | 74 |
| grounded-ai/agent-decision-3b | adapter | — | — | — | other | grounded-ai | 0 | 72 |
| RepublicOfKorokke/d1-3B-oQ8e | quantized | — | 3.12B | M | — | RepublicOfKorokke | 0 | 66 |
| rty813/d1-3B-bf16 | finetune | -bf16 | — | — | apache-2.0 | rty813 | 0 | 61 |
| RepublicOfKorokke/d1-3B-oQ8e-fp16 | quantized | -fp16 | 3.12B | M | — | RepublicOfKorokke | 0 | 53 |
| smdesai/d1-3B-CoreAI | quantized | — | — | — | other | smdesai | 1 | 0 |
| tkirsten/d1-3B-radiance-BF16 | finetune | -BF16 | — | — | other | tkirsten | 1 | 0 |
| huveewomg/d1-3B-ONNX | quantized | — | — | — | other | huveewomg | 1 | 0 |
| Vishva007/d1-3B-W4A16-AutoRound | quantized | -W4A16 | 1.08B | S | other | Vishva007 | 0 | 50 |
| johannhartmann/webnav-crawler-d1-3b-lora | adapter | — | — | — | other | johannhartmann | 1 | 0 |
| blackOrder/Ordivo-v1 | adapter | — | — | — | other | blackOrder | 0 | 35 |
| yannelli/d1-3b-cli-command-risk | finetune | — | 3.12B | M | other | yannelli | 0 | 33 |
| myroslavtryhubets/d1-3b-NVFP4 | quantized | NVFP4 | 3.12B | M | other | myroslavtryhubets | 0 | 31 |
| WasamiKirua/liquid-d1-3b-memory-gate | finetune | — | 3.12B | M | other | WasamiKirua | 0 | 29 |
| yannelli/d1-3b-cli-command-risk-v1.5 | finetune | — | 3.12B | M | other | yannelli | 0 | 13 |
| dotimothy/sima-d1 | finetune | — | — | — | other | dotimothy | 0 | 0 |
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 · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-14 14:07 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.
