Skip to content
AI Atlas
Open model frontier

Open models

Every canonical model whose weights can be downloaded, described by measurable properties — licence permissions, parameters, context, release, benchmark ranks, providers, estimated local fit. Openness is a set of observed dimensions here, never a score.

950 downloadable models

Openness explorer

Categories and licence permissions

Left: the downloadable universe by openness category. Right: what the licences on this page allow, counted per category — allowed · restricted · unknown.

Top licences: Apache License 2.0 (275) · Other (unclassified licence) (115) · MIT License (92) · Creative Commons Attribution-NonCommercial 4.0 (22) · Apple Sample Code / ML Research License (15) · Llama 3.2 Community License (12)

Category · this pageCommercial useRedistributionDerivativesHosting
Open source n=0
Open weights n=4938011380113801136013
restricted-weights n=1100100100100

Counts read: allowed · restricted · unknown / unclassified — over the 50 models on this page, from each licence's stated terms (ontology).

Downloadable models

950 models · sorted by best benchmark rank

Permissions, in order:commercial use· redistribution· derivatives· hosting allowed restricted unknown

ModelLicence · permissionsParamsContextReleaseBest resultsProvidersFit 64 GB @4bit · 128 GB @8bit Estimated
music-medium-100kOpen weightsStanford CRFMApache-2.05 Jun 20230
music-medium-800kOpen weightsStanford CRFMApache-2.05 Jun 20230
music-small-100kOpen weightsStanford CRFMApache-2.04 Jun 20230
music-small-800kOpen weightsStanford CRFMApache-2.04 Jun 20230
music-small-ar-800kOpen weightsStanford CRFMApache-2.05 Jun 20230
music-small-ar-inter-100kOpen weightsStanford CRFMApache-2.05 Jun 20230
nemotron-3.5-asr-streaming-0.6bOpen weightsNVIDIAOther638M15 May 20260✓ fits0.9 GB4bit✓ fits1.2 GB8bit
nvidia/llama-nemotron-rerank-1b-v2Open weightsNVIDIAOther1.24B16 Oct 20250✓ fits1.2 GB4bit✓ fits1.9 GB8bit
oOpen weightswunclassified0
olmOCR-2-7B-1025Open weightsAllen Institute for AIApache-2.08.29B6 Oct 20250✓ fits5.3 GB4bit✓ fits10.0 GB8bit
orpheus-3b-0.1-ftOpen weightsCanopy LabsApache-2.03.78B17 Mar 20250✓ fits2.7 GB4bit✓ fits4.8 GB8bit
orpheus-3b-0.1-pretrainedOpen weightsCanopy LabsApache-2.03.78B4 Mar 20250✓ fits2.7 GB4bit✓ fits4.8 GB8bit
owlv2-base-patch16-ensembleOpen weightsGoogleApache-2.0155M13 Oct 20230✓ fits0.6 GB4bit✓ fits0.7 GB8bit
parakeet-ctc-1.1bOpen weightsNVIDIACC-BY-4.01.06B28 Dec 20230✓ fits1.1 GB4bit✓ fits1.7 GB8bit
parakeet-tdt-0.6b-v2Open weightsNVIDIAunclassified0
parakeet-tdt-0.6b-v3Open weightsNVIDIACC-BY-4.0627.1M4 Aug 20250✓ fits0.9 GB4bit✓ fits1.2 GB8bit
perplexity-ai/llama-cpp-linux-binariesOpen weightsPerplexity AIunclassified17 Jul 20260
phi-2Open weightsMicrosoftMIT2.78B13 Dec 20230✓ fits2.1 GB4bit✓ fits3.7 GB8bit
pplx-computer-qwen-3-8-27b-dflash2-20260824Open weightsPerplexity AIOther18.8B25 Aug 20260✓ fits11.3 GB4bit✓ fits22.1 GB8bit
pplx-computer-vllm-dflash2Open weightsPerplexity AIunclassified20 Aug 20260
pplx-embed-context-v1-0.6bOpen weightsPerplexity AIMIT596M20 Jan 20260✓ fits0.8 GB4bit✓ fits1.2 GB8bit
pplx-embed-context-v1-4bOpen weightsPerplexity AIMIT4.02B20 Jan 20260✓ fits2.8 GB4bit✓ fits5.1 GB8bit
pplx-embed-v1-0.6bOpen weightsPerplexity AIMIT596M14 Jan 20260✓ fits0.8 GB4bit✓ fits1.2 GB8bit
pplx-embed-v1-4bOpen weightsPerplexity AIMIT4.02B20 Jan 20260✓ fits2.8 GB4bit✓ fits5.1 GB8bit
pplx-embed-v1-late-0.6bOpen weightsPerplexity AIMIT595.8M13 Mar 20260✓ fits0.8 GB4bit✓ fits1.2 GB8bit
pplx-pii-maskingOpen weightsPerplexity AIMIT596.1M16 Jul 20260✓ fits0.8 GB4bit✓ fits1.2 GB8bit
pplx-pii-masking-vllm-tmpOpen weightsPerplexity AIOther596.1M2 Aug 20260✓ fits0.8 GB4bit✓ fits1.2 GB8bit
pplx-qwen-3-8-27b-dflash2-20260819Open weightsPerplexity AIOther18.8B19 Aug 20260✓ fits11.3 GB4bit✓ fits22.1 GB8bit
privacy-filterOpen weightsOpenAIApache-2.01.4B17 Apr 20260✓ fits1.3 GB4bit✓ fits2.1 GB8bit
pythia-1.4bOpen weightsEleutherAIApache-2.01.52B9 Feb 20230✓ fits1.4 GB4bit✓ fits2.3 GB8bit
pythia-14mOpen weightsEleutherAIApache-2.014.1M24 Feb 20260✓ fits0.5 GB4bit✓ fits0.5 GB8bit
pythia-14m-dedupedOpen weightsEleutherAIApache-2.039.2M19 Jul 20230✓ fits0.5 GB4bit✓ fits0.6 GB8bit
pythia-160mOpen weightsEleutherAIApache-2.0212.7M8 Feb 20230✓ fits0.6 GB4bit✓ fits0.7 GB8bit
pythia-160m-dedupedOpen weightsEleutherAIApache-2.0212.7M8 Feb 20230✓ fits0.6 GB4bit✓ fits0.7 GB8bit
pythia-160m-seed1Open weightsEleutherAIApache-2.0212.7M15 Feb 20230✓ fits0.6 GB4bit✓ fits0.7 GB8bit
pythia-160m-seed2Open weightsEleutherAIApache-2.0212.7M15 Feb 20230✓ fits0.6 GB4bit✓ fits0.7 GB8bit
pythia-160m-seed3Open weightsEleutherAIApache-2.0212.7M15 Feb 20230✓ fits0.6 GB4bit✓ fits0.7 GB8bit
pythia-1bOpen weightsEleutherAIApache-2.01.08B10 Mar 20230✓ fits1.1 GB4bit✓ fits1.7 GB8bit
pythia-2.8bOpen weightsEleutherAIApache-2.02.91B13 Feb 20230✓ fits2.2 GB4bit✓ fits3.9 GB8bit
pythia-410mOpen weightsEleutherAIApache-2.0506M13 Feb 20230✓ fits0.8 GB4bit✓ fits1.1 GB8bit
pythia-410m-dedupedOpen weightsEleutherAIApache-2.0506M13 Feb 20230✓ fits0.8 GB4bit✓ fits1.1 GB8bit
pythia-6.9bOpen weightsEleutherAIApache-2.06.99B14 Feb 20230✓ fits4.5 GB4bit✓ fits8.5 GB8bit
pythia-70mOpen weightsEleutherAIApache-2.095.6M13 Feb 20230✓ fits0.6 GB4bit✓ fits0.6 GB8bit
pythia-70m-dedupedOpen weightsEleutherAIApache-2.095.6M13 Feb 20230✓ fits0.6 GB4bit✓ fits0.6 GB8bit
resnet-50Open weightsMicrosoftApache-2.025.6M16 Mar 20220✓ fits0.5 GB4bit✓ fits0.5 GB8bit
scibert_scivocab_uncasedOpen weightsAllen Institute for AIunclassified2 Mar 20220
sd-turboOpen weightsStability AIunclassified865.9M27 Nov 20230✓ fits1.0 GB4bit✓ fits1.5 GB8bit
sd-vae-ft-emaOpen weightsStability AIMIT83.7M13 Oct 20220✓ fits0.6 GB4bit✓ fits0.6 GB8bit
sd-vae-ft-mseOpen weightsStability AIMIT83.7M13 Oct 20220✓ fits0.6 GB4bit✓ fits0.6 GB8bit
sd-x2-latent-upscalerrestricted-weightsStability AIOpenRAIL++-M372.6M3 Feb 20230✓ fits0.7 GB4bit✓ fits0.9 GB8bit

Best results = rank inside each benchmark's primary comparability group; several dimensions are shown side by side and never combined. Fit columns are estimates (64 GB device at 4-bit, 128 GB at 8-bit, 8K context) — method on /methodology. unclassified = the raw licence label is not yet mapped in the ontology.

MethodUniverse = canonical models whose weights can be downloaded (open-weights, open-source, restricted-weights). hardware_fit values are ESTIMATES (see /methodology); best_results are ranks inside each benchmark's primary comparability group — no composite score. /methodology