Marin
3 models across 1 family · Latest: Marin 32B Base (2025-10)
Marin's portfolio covers 3 active models across 1 current family, spanning long context. Open a model detail page to compare provider routes and sourced benchmarks.
Covers 1 workload area across 3 active tracked models; last verified 2026-06-29.
Use it for
- Teams evaluating long context across this lab's releases
- Comparing model families before committing to a flagship
- Migration and pricing follow-ups across 3 tracked models
Do not use it for
- Choosing a hosting provider without opening a model page for price ladders
Active models
3
Current models from this lab, excluding deprecated ones
Active families
1
Current model families from this lab
Open catalog
3 open
2 open source / 1 open weights
Lowest output price
Not tracked
No provider output pricing linked yet
Latest dated release
2025-10-25
Marin 32B Base
Freshness
2026-06-29
Researched 3d ago
Information
Links
WebsiteRelease cadence
Showing 3 recent dated releases (full timeline below). Latest: Marin 32B Base (2025-10-25).
Where this lab wins
- Long-context: 1 tracked model with context-token or InfiniteBench-class signal.
Flagship quality / price signal
Flagship: Marin 8B Instruct (best sourced coding quality-per-dollar in this portfolio).
Quality-per-dollar unavailable for this flagship — benchmark coverage or output token pricing is still missing.
Marin is an American AI research organization founded in 2024. Marin ships 1 model family totaling 3 models, with the most recent release Marin 32B Base in 2025-10. Notable families include Marin. Use it as a stable reference for lab background, release coverage, and follow-up model pages as they are added. Researchers and evaluators can scan counts, links, release history, and source. View official API endpoints, benchmark performance, and coding/agent fit for every Marin model.
About
Marin is an open-source initiative by the Marin Community in close collaboration with Stanford University's Center for Research on Foundation Models (CRFM). Focused on building foundation models with emphasis on reproducibility, interpretability, and safety evaluation. Marin models use activation steering and refusal mechanisms for safer outputs, and support research in mechanistic interpretability.
Featured models
| Model | Released | Context | Input price ($/1M) | Output price ($/1M) | License | Openness |
|---|---|---|---|---|---|---|
| Marin 32B Base | 2025-10-25 | 4k | - | - | Apache 2.0 | Open source |
| Marin 8B Instruct | 2025-09-01 | 128k | - | - | Open Weights | Open weights |
| Marin 8B Base | 2025-05-15 | 4k | - | - | Apache 2.0 | Open source |
Model families
Recent releases
- Marin 32B Base- 2025-10-25
- Marin 8B Instruct- 2025-09-01
- Marin 8B Base- 2025-05-15
Top comparisons
- Llama 3.1 Nemotron Nano 8B v1 vs Marin 8B Instruct44
- Llama 3.1 Nemotron Nano VL 8B v1 vs Marin 8B Instruct28
- Llama 3.1 NemoGuard 8B Topic Control vs Marin 8B Instruct17
- Together AI - Llama 3 8B Lite vs Marin 8B Instruct16
- Llama Guard 4 12B vs Marin 8B Instruct11
- Marin 8B Instruct vs MiniCPM 2B10
- Hunyuan Hy3 Preview vs Marin 8B Instruct8
- Llama 2 7B vs Marin 8B Instruct8
FAQ
Who founded Marin and when?
Marin was founded in 2024 and is associated with Stanford, California, United States.
What models has Marin released?
Marin ships 3 models across 1 family: Marin.
Is Marin's technology open source?
All tracked Marin models are open-weight or open-source.
Where is Marin headquartered?
Marin is headquartered in Stanford, California, United States.
How can I access Marin's models?
Marin's models are available via NVIDIA NIM.
Explore related pages
Last reviewed: 2026-06-29. Data sourced from public lab announcements and provider documentation.