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

Open sourceWeights and code published under an OSI-approved licence that allows commercial use and derivatives.

1

Open weightsWeights downloadable under a permissive or Creative Commons licence allowing commercial use; code or data may be missing.

869

Restricted weightsWeights downloadable, but the licence restricts commercial use, hosting, derivatives or field of use (community, research and RAIL licences).

80

Downloadable totalUniverse = 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.

950

New in 30 daysDownloadable models with a release date inside the last 30 days (release_date claim; models without one are not counted).

269

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=501004010040100409041
restricted-weights n=0————

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
Step 3.5 FlashOpen weightsStepFun · text—unclassified—262.1K29 Jan 2026tau2-bench#28terminal-bench#71ifbench#821from $0.30 out——
deepseek-v4-flash-0420-non-reasoningOpen weightsDeepSeek—unclassified—1M—tau2-bench#28terminal-bench#65artificial-analysis-intelligence-index#1280——
Mistral Medium 3.5Open weightsMistral AI · image, textMIT-Modified—256K30 Apr 2026tau2-bench#31terminal-bench#67ifbench#681from $3.75 out——
Qwen3.6 27BOpen weightsQwen · image, text, videoApache-2.027.8B262.1K21 Apr 2026tau2-bench#31livebench-data-analysis#42livebench-coding#462from $2 out✓ fits16.5 GB4bit✓ fits32.5 GB8bit
deepseek-v4-flashOpen weightsDeepSeek—unclassified—1M—artificial-analysis-intelligence-index#31gpqa-diamond#35scicode#360——
deepseek-v4-pro-0424-highOpen weightsDeepSeek—unclassified—1M—tau2-bench#31terminal-bench#33gpqa-diamond#370——
MiMo-V2.5Open weightsXiaomi · audio, image, text, video—unclassified—1.05M22 Apr 2026terminal-bench#33mmmu-pro#40tau2-bench#541from $0.28 out——
command-a-plusOpen weightsCohere—unclassified—192K—ifbench#34scicode#89tau2-bench#970——
gemma-4-12BOpen weightsGoogle—unclassified—256K—ifbench#36mmmu-pro#72terminal-bench#1260——
gpt-oss-120bOpen weightsOpenAI · textApache-2.0116.8B131.1K4 Aug 2025swe-bench-verified#36aider-polyglot#36aider-polyglot-well-formed#584from $0.17 out✗ too large67.7 GB4bit✗ too large134.8 GB8bit
motif-3Open weightsMotif Technologies—unclassified—262.1K—artificial-analysis-intelligence-index#36humanitys-last-exam#47scicode#720——
Qwen3 32BOpen weightsQwen · textApache-2.032.8B131.1K27 Apr 2025aider-polyglot#37aider-polyglot-well-formed#57scicode#1001from $0.28 out✓ fits19.3 GB4bit✓ fits38.2 GB8bit
mimo-v2-0206Open weightsXiaomi—unclassified—256K—tau2-bench#38ifbench#48terminal-bench#790——
tri-21b-think-previewOpen weightsTrillion Labs—unclassified—32K—tau2-bench#38ifbench#158terminal-bench#2370——
Inkling SmallOpen weightsThinking Machines · audio, image, text—unclassified—1.05M30 Jul 2026scicode#39gpqa-diamond#50mmmu-pro#551from $1.2 out——
Hy3Open weightsTencent · textApache-2.0298.8B262.1K2 Jul 2026scicode#43gpqa-diamond#47humanitys-last-exam#641from $0.528 out✗ too large172.3 GB4bit✗ too large344.1 GB8bit
k2-horizon-375b-a23bOpen weightsMBZUAI Institute of Foundation Models—unclassified—524.3K—artificial-analysis-intelligence-index#43humanitys-last-exam#67scicode#680——
Qwen3.5-35B-A3BOpen weightsQwen · image, text, video—unclassified—262.1K25 Feb 2026ifbench#44tau2-bench#62mmmu-pro#661from $1.25 out——
Gemma 4 26B A4BOpen weightsGoogle · image, text, videoApache-2.025.8B4B active262.1K11 Mar 2026ifbench#45mmmu-pro#73terminal-bench#991from $0.22 out✓ fits15.3 GB4bit✓ fits30.2 GB8bit
Hy3 previewOpen weightsTencent · textOther298.8B262.1K13 Apr 2026tau2-bench#45terminal-bench#65gpqa-diamond#751from $0.60 out✗ too large172.3 GB4bit✗ too large344.1 GB8bit
solar-open2-250bOpen weightsUpstage—unclassified—1.05M—scicode#46humanitys-last-exam#81artificial-analysis-intelligence-index#820——
ring-2-6-1tOpen weightsinclusionAI—unclassified—262.1K—tau2-bench#47scicode#58gpqa-diamond#860——
Qwen3.5-4BOpen weightsQwenApache-2.04.66B262.1K27 Feb 2026tau2-bench#48mmmu-pro#85terminal-bench#1260✓ fits3.2 GB4bit✓ fits5.9 GB8bit
QwQ-32BOpen weightsAlibaba Group · textApache-2.032B131.1K6 Mar 2025aider-polyglot#49aider-polyglot-well-formed#60ifbench#2100✓ fits18.9 GB4bit✓ fits37.3 GB8bit
deepseek-v4-pro-0424-non-reasoningOpen weightsDeepSeek—unclassified—1M—terminal-bench#49tau2-bench#52artificial-analysis-intelligence-index#1150——
Nemotron 3 SuperOpen weightsNVIDIA · text—unclassified—262.1K11 Mar 2026ifbench#50terminal-bench#87scicode#981from $0.40 out——
Muse Glimmer 30BOpen weightsMeta AI · image, text—unclassified—131.1K9 Aug 2026mmmu-pro#53scicode#59humanitys-last-exam#1123from $0.75 out——
nex-n2-proOpen weightsNex AGI—unclassified—262K—artificial-analysis-intelligence-index#55gpqa-diamond#55humanitys-last-exam#590——
Nemotron 3 Nano 30B A3BOpen weightsNVIDIA · textOther31.6B3B active262.1K4 Dec 2025ifbench#56scicode#111terminal-bench#1491from $0.20 out✓ fits18.7 GB4bit✓ fits36.8 GB8bit
Trinity Large ThinkingOpen weightsArcee AI · text—unclassified—262.1K1 Apr 2026tau2-bench#57scicode#79terminal-bench#1121from $0.80 out——
deepseek-v3-2-specialeOpen weightsDeepSeek—unclassified—128K—terminal-bench#57gpqa-diamond#71humanitys-last-exam#800——
agnes-2-5-pro-alphaOpen weightsSapiens AI—unclassified—1M—artificial-analysis-intelligence-index#59humanitys-last-exam#62gpqa-diamond#640——
ling-2-6-1tOpen weightsinclusionAI—unclassified—262.1K—tau2-bench#59terminal-bench#79ifbench#1160——
LongCat 2.0Open weightsMeituan · text—unclassified—1.05M20 Jul 2026humanitys-last-exam#60scicode#97artificial-analysis-intelligence-index#1231from $1.2 out——
MiniMax M2.1Open weightsMiniMax · textOther228.7B204.8K20 Dec 2025ifbench#63tau2-bench#81terminal-bench#871from $1.2 out✗ too large132.0 GB4bit✗ too large263.5 GB8bit
apriel-v1-6-15b-thinkerOpen weightsServiceNow—unclassified—128K—ifbench#65tau2-bench#118terminal-bench#1470——
minicpm-v4-6-1-3bOpen weightsOpenBMB—unclassified—262K—tau2-bench#68mmmu-pro#144terminal-bench#2760——
hyperclova-x-seed-think-32bOpen weightsNaver—unclassified—128K—tau2-bench#69terminal-bench#156artificial-analysis-intelligence-index#2160——
DeepSeek V3Open weightsDeepSeek · text—unclassified684.5B163.8K25 Dec 2024terminal-bench#71scicode#101tau2-bench#1041from $1.03 out✗ too large394.1 GB4bit✗ too large787.7 GB8bit
Qwen3.5-9BOpen weightsQwen · image, text, videoApache-2.09.65B262.1K27 Feb 2026tau2-bench#72mmmu-pro#73ifbench#802from $0.15 out✓ fits6.0 GB4bit✓ fits11.6 GB8bit
Ling 3.0 FlashOpen weightsinclusionAI · text—unclassified—262.1K23 Jul 2026scicode#74artificial-analysis-intelligence-index#78gpqa-diamond#911from $0.063 out——
k-exaone-2-0-0803Open weightsLG AI Research—unclassified—262.1K—scicode#74gpqa-diamond#120artificial-analysis-intelligence-index#1230——
DeepSeek V3.1 TerminusOpen weightsDeepSeek · text—unclassified—163.8K22 Sept 2025terminal-bench#75scicode#91ifbench#1141from $1 out——
Qwen3 235B A22B Instruct 2507Open weightsQwen · text—unclassified—262.1K21 Jul 2025scicode#76ifbench#139tau2-bench#1391from $0.35 out——
a-x-k2Open weightsSK Telecom—unclassified—262.1K—humanitys-last-exam#76scicode#78gpqa-diamond#860——
DeepSeek V3.2 ExpOpen weightsDeepSeek · text—unclassified—163.8K29 Sept 2025terminal-bench#79ifbench#127gpqa-diamond#1421from $0.41 out——
North Mini Code (free)Open weightsCohereApache-2.030.5B256K5 Jun 2026terminal-bench#79scicode#87ifbench#1111✓ fits18.0 GB4bit✓ fits35.5 GB8bit
Qwen3 VL 235B A22B InstructOpen weightsQwen · image, text—unclassified—262.1K23 Sept 2025mmmu-pro#79ifbench#118tau2-bench#1371from $1.9 out——
ling-2-6-flashOpen weightsinclusionAI—unclassified—262.1K—tau2-bench#79ifbench#112terminal-bench#1160——
exaone-4-5-33bOpen weightsLG AI Research—unclassified—262.1K—mmmu-pro#83tau2-bench#105ifbench#1080——

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