An uncensored community variant of Swift 1.5 Qwen3.8-27B in native NInfer format, with mixed NVFP4/FP8 weights and embedded DFlash2 speculative decoding. This is a third-party release, not an official UkisAI or OrcaRouter checkpoint. SHA-256: d094126721091481d8fa3b34f780eb68fcaac61ef0750b698cb5076b6b2e84b1 Use a compatible NInfer build that supports Qwen3.8-27B NVFP4 and DFlash2. See the upstream project for installation and supported hardware. This file is not GGUF and cannot be loaded directly with Transformers or llama.cpp. Add --vision for image input. Increase --max-context as available VRAM allows; the example starts at 8,192 tokens. DFlash2 is already embedded, so no separate draft…
SAVRN Model Hub
Open-Weight Models
An open-weight model is an AI model whose trained weights are published for anyone to download. The weights are what the model learned in training. With a copy of them you can run the model on hardware you control and train it further on your own data.
Open weights are not the same as open source. Many publishers release the weights without the training data or code, and the license sets what you may do with the model. This library puts each model's full card, architecture, files, license and published evaluations on one page.
12,283Models
3,510Datasets
290Papers
7,436Publishers
12,632Sourced relationships
Updated 2026-10-22 · How the library is built
12,283 models, sorted by most downloaded.
static quants of https://huggingface.co/d0xin/Swift-1.5-Qwen3.8-27B-Uncensored-BF16 For a convenient overview and download list, visit our model page for this model. weighted/imatrix quants are available at https://huggingface.co/mradermacher/Swift-1.5-Qwen3.8-27B-Uncensored-BF16-i1-GGUF If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on the matter…
weighted/imatrix quants of https://huggingface.co/d0xin/Swift-1.5-Qwen3.8-27B-Uncensored-BF16 For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/Swift-1.5-Qwen3.8-27B-Uncensored-BF16-GGUF This is a vision model - mmproj files (if any) will be in the static repository. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here…
GGUF quants of ajgazin/Swift-1.5-Qwen3.8-27B-Uncensored-MTP, an abliterated Swift 1.5 Qwen3.8-27B (UkisAI's reasoning-efficient fine-tune of Qwen3.8-27B). For vLLM and size, with Unsloth's importance matrix. - MTP head included in every main GGUF, for self-speculative decoding in llama.cpp. Each quant is one file, Swift-1.5-Qwen3.8-27B-Uncensored-Dynamic-MTP-.gguf; the BF16 is Swift-1.5-Qwen3.8-27B-Uncensored-MTP-BF16.gguf. --spec-type draft-mtp needs a llama.cpp build with MTP support for qwen35. The MTP head loads from the main GGUF; there is no separate draft file. Sampling, as for Swift and Qwen: temperature 1.0, topp 0.95, topk 20, minp 0. The model thinks before answering by default.…
basemodel: - ajgazin/Swift-1.5-Qwen3.8-27B-Uncensored-MTP
weighted/imatrix quants of https://huggingface.co/ajgazin/Swift-1.5-Qwen3.8-27B-Uncensored-MTP For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/Swift-1.5-Qwen3.8-27B-Uncensored-MTP-GGUF This is a vision model - mmproj files (if any) will be in the static repository. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here…
NVFP4 checkpoint of an abliterated Swift 1.5 Qwen3.8-27B (UkisAI's reasoning-efficient fine-tune of Qwen3.8-27B). For vLLM and SGLang. GGUFs for llama.cpp: (measured on the BF16 source). - Swift's own NVFP4 recipe, unmodified, from ukisai/Swift-Qwen3.8-27B-NVFP4, calibrated with NVIDIA ModelOpt. The module split matches UkisAI's Swift 1.5 NVFP4 exactly. - MTP head and vision tower in BF16, bit-identical to the source. 21.9 GB, NVIDIA ModelOpt mixed-precision format. Needs a vLLM with ModelOpt mixed-precision support. No --quantization flag. Sampling, as for Swift and Qwen: temperature 1.0, topp 0.95, topk 20, minp 0. The model thinks before answering by default. Same format, recipe, module…
Swift 1.5 Qwen3.8-27B is UkisAI's reasoning-efficient derivative of Qwen3.8-27B. It uses 58.5% fewer thinking tokens while scoring 0.35% higher than the base, for a 1.95× speed-up on several tasks. Swift 1.5 is a direct upgrade from Swift 1.0, our model with 350k+ downloads, delivering stronger overall performance than both base and Swift 1.0 in various tasks, especially coding and agentic, while using fewer thinking tokens. We accomplished that by scaling up the post-training (RL and OPD) from the previous version. We gave base Qwen3.8-27B and Swift 1.5 27B the same prompt: Base Qwen3.8-27B took 104.6 minutes to build its game. Swift 1.5 took 11.39 minutes. We made Swift efficient by…
static quants of https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b For a convenient overview and download list, visit our model page for this model. weighted/imatrix quants seem not to be available (by me) at this time. If they do not show up a week or so after the static ones, I have probably not planned for them. Feel free to request them by opening a Community Discussion. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality…
An abliterated build of ukisai/Swift-1.5-Qwen3.8-27b, UkisAI's reasoning-efficient derivative of Qwen3.8-27B. Refusal directions are removed by directional ablation; nothing else is retrained. identical to the source's), lmhead, embedtokens, all norms, and every GatedDeltaNet recurrent gate. - Size 52G, 1,199 tensors (333 vision, 15 MTP). Produced with Heretic v2.0.0.dev0: 260 TPE trials, seed 42, directional ablation on attn.oproj (16 modules), attn.outproj (48) and mlp.downproj (64). Baseline refusal score before abliteration: 98/100. Trial 140 was selected — Pareto index 1, not index 0. Index 0 (trial 161) scored 22/100 keywords at KL 0.1091; trial 140 scores 23/100 at KL 0.0602. One…
A single Radiance container (.rad, 18.05 GiB) holding the target model and its speculative drafter, quantised for AMD RDNA4 (Radeon AI PRO R9700, gfx1201, 32 GB). It is not a Transformers/vLLM checkpoint: it runs only with the Radiance inference engine. The GPTQ source checkpoint is published separately as 1. Qwen/Qwen3.8-27B (Apache-2.0) 2. ukisai/Swift-1.5-Qwen3.8-27b (Swift Open License v1.0) 3. akumaburn/Swift-1.5-Qwen3.8-27b-heretic (abliterated, BF16) 4. GPTQ to MXFP4 (llm-compressor, MXFP4A16, static act-order, 89 calibration samples, lmhead/vision/MTP excluded — gptq-recipe.yaml) 5. rad-convert (Radiance 1.2.3) with the DFlash2 drafter incoai/Qwen3.8-27B-DFlash2 (Apache-2.0) merged…
An INT8 W8A8 quantization of (the abliterated BF16 build of ukisai/Swift-1.5-Qwen3.8-27b), for fast serving on GPUs where native FP8 is unavailable or undesirable. native CUTLASS INT8 tensor-core path. No activation scales are stored on disk; they are computed per token at runtime. vision tower (333 tensors), GatedDeltaNet recurrent gates, all norms. These are FP16, not the source's BF16 — the 16-bit residual precision of this build is float16 throughout (config.json: "dtype": "float16"). with the MTP drafter re-expressed and ablated in the rotated basis. The rotation was verified before calibration by an fp64 whole-graph equivalence check (all 19 paths within 1.6–2.5e-4; every tensor…
A 18G INT4 build of for single-user decoding and low-VRAM deployment. (Marlin on Ampere). - Unrotated, so the DFlash2 drafter works. GatedDeltaNet gates, all norms. 4-bit costs 4.5× the rotated W8A8's quantization divergence (0.0264) and 1.5× the unrotated W8A8's (0.0791). Choose this build when the 18G footprint or single-user decode speed matters more than fidelity. Produced with Heretic v2.0.0.dev0: 260 TPE trials, seed 42, directional ablation on attn.oproj (16 modules), attn.outproj (48) and mlp.downproj (64). Baseline refusal score before abliteration: 98/100. Trial 140 was selected — Pareto index 1, not index 0. Index 0 (trial 161) scored 22/100 keywords at KL 0.1091; trial 140…
Three custom mixed-tensor quantizations of ukisai/Swift-Qwen3.8-27b, a fine-tune of Qwen/Qwen3.8-27B — the dense 27B multimodal model with hybrid GatedDeltaNet + Gated Attention and a Multi-Token Prediction (MTP) head. Built with llama.cpp 58367713a, using a purpose-built importance matrix and per-tensor quantization overrides. All three are designed around the 16 GB VRAM budget of consumer GPUs such as the AMD Radeon RX 9070 XT, and all keep the MTP head embedded so speculative decoding works out of the box — no second file to download or load. Both quants follow the same philosophy: protect what the model cannot afford to lose, and pay for it out of the FFN, which is two-thirds of the…
An uncensored community variant of Swift Qwen3.8-27B in native NInfer format, with mixed NVFP4/FP8 weights and embedded DFlash2 speculative decoding. This is a third-party release, not an official UkisAI or OrcaRouter checkpoint. SHA-256: b57bb82f6c43fa4bba3bcb883f45ccff12b81fd461502a37c79d78c3ba7a979b Use a compatible NInfer build that supports Qwen3.8-27B NVFP4 and DFlash2. See the upstream project for installation and supported hardware. This file is not GGUF and cannot be loaded directly with Transformers or llama.cpp. Add --vision for image input. Increase --max-context as available VRAM allows; the example starts at 8,192 tokens. DFlash2 is already embedded, so no separate draft…
static quants of https://huggingface.co/vwdubb/Swift-Qwen3.8-27B-Uncensored-MTP-Terse-Coder For a convenient overview and download list, visit our model page for this model. weighted/imatrix quants are available at https://huggingface.co/mradermacher/Swift-Qwen3.8-27B-Uncensored-MTP-Terse-Coder-i1-GGUF If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on the matter…
weighted/imatrix quants of https://huggingface.co/vwdubb/Swift-Qwen3.8-27B-Uncensored-MTP-Terse-Coder For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/Swift-Qwen3.8-27B-Uncensored-MTP-Terse-Coder-GGUF This is a vision model - mmproj files (if any) will be in the static repository. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality…
An INT8 W8A8 quantization of (the abliterated BF16 build of a merged LoRA finetune of Qwen/Qwen3.8-27B), for fast serving on GPUs where native FP8 is unavailable or undesirable. int-quantized, native CUTLASS INT8 tensor-core path. embedtokens, the vision tower, GatedDeltaNet recurrent gates, all norms. Note these are FP16, not the source's BF16 — the 16-bit residual precision of this build is float16 throughout (config.json: "dtype": "float16"). (llm-compressor), with the MTP drafter ablated in the rotated basis. The MTP head is unquantized but is not byte-identical to the BF16 build's: it carries the same residual-stream rotation and RMSNorm fold as the rest of the model, so the two heads…
This is a quantized MLX pack of ukisai/Swift1.5-Qwen3.8-Flash-Next — UkisAI's reasoning-efficient finetune of Qwen3.8-Flash-Next — produced by applying the MTPLX "Optimized Speed" quantization recipe from Youssofal/Qwen3.8-Flash-Next-MTPLX-Optimized-Speed to the Swift-1.5 weights. It was forged locally with mtplx forge (MTPLX 2.12.2, qwen4-exp lane) from source revision 0bd4fe22431372cdad1979267d3ab45aa7e6150a, full-load verified before stamping, and runs with MTP draft-acceptance active and the n-gram table offloaded to SSD — the same target class as the original pack: M5 Ultra 96 GB or M5 Max 128 GB Macs with fast NVMe storage. Note vs the original Youssofal pack: this build runs qsa8bit…
Model Collections
Hand-picked starting points, each with the reason it exists.
Collection · 4 entries
Embedding models for retrieval
Sentence and document embedding models used to build retrieval systems. Dimension and sequence length matter more than size here, and both come from the publisher.
Collection · 4 entries
Models that fit on one accelerator
Models whose publisher-reported parameter count puts them within reach of a single accelerator at common precisions. Memory needed depends on precision and serving configuration, so treat the parameter count as the starting point, not the answer.
Collection · 6 entries
Open-weight text models worth knowing
Widely used open-weight language models, chosen because each one is a distinct family rather than a variant of the one above it. Selection, not a ranking.
Collection · 3 entries
Speech and audio models
Recognition and synthesis models, grouped so the two directions are easy to compare.
Open-Weight Models Explained
What is an open-weight model?
An AI model whose trained weights are published for anyone to download, so it can be run, tested and fine-tuned on hardware the user controls.
Is an open-weight model the same as open source?
Not always. Open weights means the trained model can be downloaded. Open source usually also means the training code and data are available and the license allows broad reuse. Many open-weight models release the weights only.
Can I use an open-weight model commercially?
It depends on the license. Apache 2.0 and MIT allow commercial use. Other licenses limit it, for example to non-commercial use or below a set number of users. Every model page here shows its license.
How much memory does an open-weight model need?
About two bytes per parameter at 16-bit precision, so a 7-billion-parameter model needs roughly 14 GB for its weights, plus memory for the context it processes. Each model page lists its parameter count and the size of its files.
Related SAVRN Research
The hub sits beside SAVRN's market data and infrastructure research: what models cost to run, and what it takes to run them.
SAVRN Index
What open models cost to run
The same open-weight model priced by every host that serves it, per million tokens.
Research Hub
Data center trackers and maps
Moratoriums, permits, power, water and capital behind the facilities that run these models.
Method
How the Model Hub is built
Sources, evidence labels, refresh behaviour, and the limits of every comparison here.