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=4932017320173201732017
restricted-weights n=1010100100010

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
SmolVLM2-256M-Video-InstructOpen weightsHugging FaceApache-2.0256.5M11 Feb 20250✓ fits0.7 GB4bit✓ fits0.8 GB8bit
SmolVLM2-500M-Video-InstructOpen weightsHugging FaceApache-2.0507.5M11 Feb 20250✓ fits0.8 GB4bit✓ fits1.1 GB8bit
Stable-DiffCoder-8B-BaseOpen weightsByteDanceMIT8.25B15 Jan 20260✓ fits5.3 GB4bit✓ fits10.0 GB8bit
Stable-DiffCoder-8B-InstructOpen weightsByteDanceMIT8.25B15 Jan 20260✓ fits5.3 GB4bit✓ fits10.0 GB8bit
Step-3.5-Flash-REAP-121B-A11BOpen weightsCerebras SystemsOther121B11B active25 Feb 20260✗ too large70.0 GB4bit✗ too large139.6 GB8bit
Step-3.5-Flash-REAP-149B-A11BOpen weightsCerebras SystemsOther149.4B11B active25 Feb 20260✗ too large86.4 GB4bit✗ too large172.3 GB8bit
StripedHyena-Hessian-7BOpen weightsTogether AIApache-2.07.65B21 Nov 20230✓ fits4.9 GB4bit✓ fits9.3 GB8bit
StripedHyena-Nous-7BOpen weightsTogether AIApache-2.07.65B4 Dec 20230✓ fits4.9 GB4bit✓ fits9.3 GB8bit
StyleTTS2-LJSpeechOpen weightsyl4579unclassified0
SynLogic-32BOpen weightsMiniMaxMIT32.8B30 May 20250✓ fits19.3 GB4bit✓ fits38.2 GB8bit
SynLogic-7BOpen weightsMiniMaxMIT7.62B3 Jun 20250✓ fits4.9 GB4bit✓ fits9.3 GB8bit
SynLogic-Mix-3-32BOpen weightsMiniMaxMIT32.8B30 May 20250✓ fits19.3 GB4bit✓ fits38.2 GB8bit
TRELLIS-image-largeOpen weightsMicrosoftMIT2 Dec 20240
TRELLIS.2-4BOpen weightsMicrosoftMIT4B1 Dec 20250✓ fits2.8 GB4bit✓ fits5.1 GB8bit
Thomson-1.0-SmallOpen weightsthomsonreutersunclassified0
Titan-text-embeddings-v2Open weightsAmazon Web ServicesOther30 Apr 20240
TripoSROpen weightsStability AIMIT29 Feb 20240
UI-Mate-9BOpen weightsTencentunclassified0
UI-TARS 7BOpen weightsByteDance · image, textApache-2.08.29B128K16 Apr 20251from $0.20 out✓ fits5.3 GB4bit✓ fits10.0 GB8bit
UI-TARS-2B-SFTOpen weightsByteDanceApache-2.02.44B20 Jan 20250✓ fits1.9 GB4bit✓ fits3.3 GB8bit
UI-TARS-72B-DPOOpen weightsByteDanceApache-2.073.4B22 Jan 20250✓ fits42.7 GB4bit✓ fits84.9 GB8bit
UI-TARS-7B-DPOOpen weightsByteDanceApache-2.08.29B22 Jan 20250✓ fits5.3 GB4bit✓ fits10.0 GB8bit
UI-TARS-7B-SFTOpen weightsByteDanceApache-2.08.29B20 Jan 20250✓ fits5.3 GB4bit✓ fits10.0 GB8bit
UncensoredOpen weightsVenice · textunclassified128K9 Jul 20251from $0.90 out
Unlimited-OCROpen weightsBaiduMIT3.34B19 Jun 20260✓ fits2.4 GB4bit✓ fits4.3 GB8bit
UnslopNemo 12BOpen weightsTheDrummer · textunclassified1.02M8 Nov 20241from $0.40 out
VTP-Large-f16d64Open weightsMiniMaxOther731.6M16 Dec 20250✓ fits0.9 GB4bit✓ fits1.3 GB8bit
VTP-Small-f16d64Open weightsMiniMaxOther167.2M16 Dec 20250✓ fits0.6 GB4bit✓ fits0.7 GB8bit
VibeVoice-ASROpen weightsMicrosoftMIT8.67B21 Jan 20260✓ fits5.5 GB4bit✓ fits10.5 GB8bit
Voxtral Mini Transcribe RealtimeOpen weightsMistral AIApache-2.00
Voxtral SmallOpen weightsMistral AIApache-2.032K1from $0.40 out
Voxtral Small 24B 2507Open weightsMistral AI · audio, textApache-2.024.3B32.8K1 Jul 20251from $0.30 out✓ fits14.4 GB4bit✓ fits28.4 GB8bit
Voxtral-Mini-3B-2507Open weightsMistral AIApache-2.04.68B1 Jul 20250✓ fits3.2 GB4bit✓ fits5.9 GB8bit
Voxtral-Mini-4B-Realtime-2602Open weightsMistral AIApache-2.04.43B21 Jan 20260✓ fits3.0 GB4bit✓ fits5.6 GB8bit
WeMM-Embedding-2BOpen weightsTencentOther2.72B25 Aug 20260✓ fits2.1 GB4bit✓ fits3.6 GB8bit
Whisper Large V3 TurboOpen weightsOpenAIMIT808.9M1 Oct 20240✓ fits1.0 GB4bit✓ fits1.4 GB8bit
WizardLM-2 8x22BOpen weightsMicrosoft · textunclassified65.5K16 Apr 20241from $0.62 out
XYZ-Aquila-miniOpen weightsXYZAILabunclassified0
Yi-34BOpen weights01-aiunclassified0
Youtu-LLM-2BOpen weightsTencentOther1.96B31 Dec 20250✓ fits1.6 GB4bit✓ fits2.8 GB8bit
Youtu-LLM-2B-BaseOpen weightsTencentunclassified0
academic-ds-9BOpen weightsByteDanceApache-2.09.37B9 Apr 20250✓ fits5.9 GB4bit✓ fits11.3 GB8bit
aimv2-large-patch14-224-litrestricted-weightsAppleApple-AMLR436.7M20 Nov 20240✓ fits0.8 GB4bit✓ fits1.0 GB8bit
alias-gpt2-small-x21Open weightsStanford CRFMApache-2.02 Mar 20220
amazon/GPT-OSS-20B-P-EAGLEOpen weightsAmazon Web ServicesApache-2.01.8B20 Feb 20260✓ fits1.5 GB4bit✓ fits2.6 GB8bit
amazon/MistralLiteOpen weightsAmazon Web ServicesApache-2.016 Oct 20230
amazon/Qwen3-Coder-30B-A3B-Instruct-P-EAGLE-long-contextOpen weightsAmazon Web ServicesApache-2.0972.9M3B active14 May 20260✓ fits1.1 GB4bit✓ fits1.6 GB8bit
amazon/gpt-oss-120b-p-eagleOpen weightsAmazon Web ServicesApache-2.01.7B9 Feb 20260✓ fits1.5 GB4bit✓ fits2.5 GB8bit
amazon/gpt-oss-120b-p-eagle-long-contextOpen weightsAmazon Web ServicesApache-2.01.7B14 May 20260✓ fits1.5 GB4bit✓ fits2.5 GB8bit
arwen-gpt2-medium-x21Open weightsStanford CRFMunclassified2 Mar 20220

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