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.
- Open source1 0%
- Open weights869 91%
- Restricted weights80 8%
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 page | Commercial use | Redistribution | Derivatives | Hosting |
|---|---|---|---|---|
| Open source n=0 | — | — | — | — |
| Open weights n=45 | 26019 | 26019 | 26019 | 26019 |
| restricted-weights n=5 | 140 | 500 | 500 | 050 |
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
| Model | Licence · permissions | Params | Context | Release | Best results | Providers | Fit 64 GB @4bit · 128 GB @8bit Estimated | |
|---|---|---|---|---|---|---|---|---|
| DeepSeek-V3.2-Exp-BaseOpen weightsDeepSeek | —unclassified | — | — | — | — | 0 | —— | |
| DeepSeek-V4Open weightsDeepSeek · text | —unclassified | 1.6T49B active | 1M | — | — | 0 | ✗ too large920.5 GB4bit✗ too large1840.5 GB8bit | |
| DeepSeek-V4-Flash-BaseOpen weightsDeepSeek | —unclassified | 292B | — | 22 Apr 2026 | — | 0 | ✗ too large168.4 GB4bit✗ too large336.3 GB8bit | |
| DeepSeek-V4-Flash-DSparkOpen weightsDeepSeek | MIT | 165.3B | — | 27 Jun 2026 | — | 0 | ✗ too large95.5 GB4bit✗ too large190.6 GB8bit | |
| DepthCrafterOpen weightsTencent | Other | — | — | 14 Sept 2024 | — | 0 | —— | |
| DepthProrestricted-weightsApple | Apple-AMLR | — | — | 3 Oct 2024 | — | 0 | —— | |
| DepthPro-hfrestricted-weightsApple | Apple-AMLR | 952M | — | 27 Nov 2024 | — | 0 | ✓ fits1.1 GB4bit✓ fits1.6 GB8bit | |
| Devstral-Small-2-24B-Instruct-2512Open weightsMistral AI | Apache-2.0 | 24B | — | 28 Nov 2025 | — | 0 | ✓ fits14.3 GB4bit✓ fits28.1 GB8bit | |
| Dragon-DocChat-Context-EncoderOpen weightsCerebras Systems | Other | — | — | 16 Aug 2024 | — | 0 | —— | |
| Dragon-DocChat-Query-EncoderOpen weightsCerebras Systems | Other | — | — | 16 Aug 2024 | — | 0 | —— | |
| ERNIE 4.5 VL 424B A47BOpen weightsBaidu · image, text | Apache-2.0 | 423.5B47B active | 123K | 28 Jun 2025 | — | 1from $1.25 out | ✗ too large244.0 GB4bit✗ too large487.6 GB8bit | |
| ERNIE-4.5-0.3B-Base-PTOpen weightsBaidu | Apache-2.0 | 360.8M | — | 28 Jun 2025 | — | 0 | ✓ fits0.7 GB4bit✓ fits0.9 GB8bit | |
| ERNIE-4.5-0.3B-PTOpen weightsBaidu | Apache-2.0 | 360.8M | — | 28 Jun 2025 | — | 0 | ✓ fits0.7 GB4bit✓ fits0.9 GB8bit | |
| ERNIE-4.5-0.3B-PaddleOpen weightsBaidu | Apache-2.0 | 360.8M | — | 29 Jun 2025 | — | 0 | ✓ fits0.7 GB4bit✓ fits0.9 GB8bit | |
| ERNIE-4.5-21B-A3B-Base-PTOpen weightsBaidu | Apache-2.0 | 21.8B3B active | — | 28 Jun 2025 | — | 0 | ✓ fits13.1 GB4bit✓ fits25.6 GB8bit | |
| ERNIE-4.5-21B-A3B-PTOpen weightsBaidu | Apache-2.0 | 21.9B3B active | — | 28 Jun 2025 | — | 0 | ✓ fits13.1 GB4bit✓ fits25.7 GB8bit | |
| ERNIE-4.5-21B-A3B-ThinkingOpen weightsBaidu | Apache-2.0 | 21.8B3B active | — | 8 Sept 2025 | — | 0 | ✓ fits13.1 GB4bit✓ fits25.6 GB8bit | |
| ERNIE-4.5-300B-A47B-Base-PTOpen weightsBaidu | Apache-2.0 | 299.5B47B active | — | 28 Jun 2025 | — | 0 | ✗ too large172.7 GB4bit✗ too large344.9 GB8bit | |
| ERNIE-4.5-300B-A47B-PTOpen weightsBaidu | Apache-2.0 | 300.5B47B active | — | 28 Jun 2025 | — | 0 | ✗ too large173.3 GB4bit✗ too large346.0 GB8bit | |
| ERNIE-4.5-300B-A47B-PaddleOpen weightsBaidu | Apache-2.0 | 300.5B47B active | — | 28 Jun 2025 | — | 0 | ✗ too large173.3 GB4bit✗ too large346.0 GB8bit | |
| ERNIE-4.5-VL-28B-A3B-PTOpen weightsBaidu | Apache-2.0 | 29.4B3B active | — | 28 Jun 2025 | — | 0 | ✓ fits17.4 GB4bit✓ fits34.3 GB8bit | |
| ERNIE-4.5-VL-28B-A3B-ThinkingOpen weightsBaidu | Apache-2.0 | 29.7B3B active | — | 7 Nov 2025 | — | 0 | ✓ fits17.6 GB4bit✓ fits34.6 GB8bit | |
| ERNIE-4.5-VL-424B-A47B-Base-PTOpen weightsBaidu | Apache-2.0 | 423.5B47B active | — | 28 Jun 2025 | — | 0 | ✗ too large244.0 GB4bit✗ too large487.6 GB8bit | |
| ERNIE-ImageOpen weightsBaidu | Apache-2.0 | 8.03B | — | 7 Apr 2026 | — | 0 | ✓ fits5.1 GB4bit✓ fits9.7 GB8bit | |
| ERNIE-Image-AesOpen weightsBaidu | Apache-2.0 | 7.94B | — | 18 May 2026 | — | 0 | ✓ fits5.1 GB4bit✓ fits9.6 GB8bit | |
| ERNIE-Image-TurboOpen weightsBaidu | Apache-2.0 | 8.03B | — | 2 Apr 2026 | — | 0 | ✓ fits5.1 GB4bit✓ fits9.7 GB8bit | |
| FLUX.1-Canny-dev-loraOpen weightsBlack Forest Labs | Other | — | — | 20 Nov 2024 | — | 0 | —— | |
| FLUX.1-Depth-dev-loraOpen weightsBlack Forest Labs | Other | — | — | 20 Nov 2024 | — | 0 | —— | |
| FLUX.1-Fill-devOpen weightsBlack Forest Labs | Other | 11.9B | — | 20 Nov 2024 | — | 0 | ✓ fits7.3 GB4bit✓ fits14.2 GB8bit | |
| FLUX.1-Kontext-devOpen weightsBlack Forest Labs | Other | 11.9B | — | 28 May 2025 | — | 0 | ✓ fits7.3 GB4bit✓ fits14.2 GB8bit | |
| FLUX.1-Krea-devOpen weightsBlack Forest Labs | Other | 11.9B | — | 7 Jul 2025 | — | 0 | ✓ fits7.3 GB4bit✓ fits14.2 GB8bit | |
| FLUX.1-devOpen weightsBlack Forest Labs | Other | 11.9B | — | 31 Jul 2024 | — | 0 | ✓ fits7.3 GB4bit✓ fits14.2 GB8bit | |
| FLUX.1-schnellOpen weightsBlack Forest Labs | Apache-2.0 | 11.9B | — | 31 Jul 2024 | — | 0 | ✓ fits7.3 GB4bit✓ fits14.2 GB8bit | |
| FLUX.2-devOpen weightsBlack Forest Labs | Other | 32.2B | — | 22 Nov 2025 | — | 0 | ✓ fits19.0 GB4bit✓ fits37.5 GB8bit | |
| FLUX.2-klein-4BOpen weightsBlack Forest Labs | Apache-2.0 | 3.88B | — | 14 Jan 2026 | — | 0 | ✓ fits2.7 GB4bit✓ fits5.0 GB8bit | |
| FLUX.2-klein-9BOpen weightsBlack Forest Labs | Other | 9.08B | — | 14 Jan 2026 | — | 0 | ✓ fits5.7 GB4bit✓ fits10.9 GB8bit | |
| FLUX.2-klein-9b-kvOpen weightsBlack Forest Labs | Other | 9.08B | — | 9 Mar 2026 | — | 0 | ✓ fits5.7 GB4bit✓ fits10.9 GB8bit | |
| FLUX.2-klein-base-4BOpen weightsBlack Forest Labs | Apache-2.0 | 3.88B | — | 14 Jan 2026 | — | 0 | ✓ fits2.7 GB4bit✓ fits5.0 GB8bit | |
| FLUX.2-klein-base-9BOpen weightsBlack Forest Labs | Other | 9.08B | — | 14 Jan 2026 | — | 0 | ✓ fits5.7 GB4bit✓ fits10.9 GB8bit | |
| FLUX.2-small-decoderOpen weightsBlack Forest Labs | Apache-2.0 | 62.4M | — | 6 Apr 2026 | — | 0 | ✓ fits0.5 GB4bit✓ fits0.6 GB8bit | |
| Fara1.5-27BOpen weightsMicrosoft | —unclassified | — | — | — | — | 0 | —— | |
| Fara1.5-4BOpen weightsMicrosoft | —unclassified | — | — | — | — | 0 | —— | |
| Fara1.5-9BOpen weightsMicrosoft | —unclassified | — | — | — | — | 0 | —— | |
| FastVLM-0.5Brestricted-weightsApple | Apple-AMLR | 758.8M | — | 25 Aug 2025 | — | 0 | ✓ fits0.9 GB4bit✓ fits1.4 GB8bit | |
| FastVLM-1.5Brestricted-weightsApple | Apple-AMLR | 1.91B | — | 25 Aug 2025 | — | 0 | ✓ fits1.6 GB4bit✓ fits2.7 GB8bit | |
| FireLLaVA-13brestricted-weightsFireworks AI | Llama-2-Community | 13.3B | — | 5 Jan 2024 | — | 0 | ✓ fits8.2 GB4bit✓ fits15.8 GB8bit | |
| Florence-2-baseOpen weightsMicrosoft | MIT | 231.6M | — | 15 Jun 2024 | — | 0 | ✓ fits0.6 GB4bit✓ fits0.8 GB8bit | |
| Florence-2-largeOpen weightsMicrosoft | MIT | 776.7M | — | 15 Jun 2024 | — | 0 | ✓ fits0.9 GB4bit✓ fits1.4 GB8bit | |
| GDN-primed-HQwen3-8B-InstructOpen weightsAmazon Web Services | Apache-2.0 | 8.5B | — | 31 Mar 2026 | — | 0 | ✓ fits5.4 GB4bit✓ fits10.3 GB8bit | |
| GKA-primed-HQwen3-32B-ReasonerOpen weightsAmazon Web Services | Apache-2.0 | 34.1B | — | 31 Mar 2026 | — | 0 | ✓ fits20.1 GB4bit✓ fits39.8 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