Magic
2 models across 1 family · Latest: LTM-2-mini (2024-08)
100M-token context window innovation
Magic's portfolio covers 2 active models across 1 current family, spanning coding and long context. Open a model detail page to compare provider routes and sourced benchmarks.
Covers 2 workload areas across 2 active tracked models; last verified 2026-05-19.
Use it for
- Teams evaluating coding and long context across this lab's releases
- Comparing model families before committing to a flagship
- Migration and pricing follow-ups across 2 tracked models
Do not use it for
- Choosing a hosting provider without opening a model page for price ladders
Active models
2
Current models from this lab, excluding deprecated ones
Active families
1
Current model families from this lab
Open catalog
0 open
0 open source / 0 open weights
Lowest output price
Not tracked
No provider output pricing linked yet
Latest dated release
2024-08-29
LTM-2-mini
Freshness
2026-05-19
Researched 60d ago
Information
Release cadence
Showing 2 recent dated releases (full timeline below). Latest: LTM-2-mini (2024-08-29).
Where this lab wins
- Coding: 2 tracked models with SWE-bench / HumanEval-style scores.
- Long-context: 2 tracked models with context-token or InfiniteBench-class signal.
Flagship quality / price signal
Flagship: LTM-1 (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.
Magic is an American AI research organization founded in 2022. 100M-token context window innovation. Magic ships 1 model family totaling 2 models, with the most recent release LTM-2-mini in 2024-08. Notable families include LTM. 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. View official API endpoints, benchmark performance, and coding/agent fit for every Magic model.
About
Magic, a trailblazing AI research company based in San Francisco, has rapidly established itself as a pivotal force in the domain of generative AI and Large Language Models (LLMs) since its inception in 2022. The company's mission is to develop safe and beneficial Artificial General Intelligence (AGI) that can tackle significant global challenges. Instead of relying solely on human efforts, Magic emphasizes the automation of AI research and code generation. This innovative approach addresses alignment issues and enhances model reliability, setting Magic apart in the crowded AI landscape. One of Magic's most remarkable contributions is its development of models capable of processing ultra-long context windows. Their pioneering LTM Net model, for example, supports a context window of five million tokens, a stark contrast to competitors like Google Gemini 1.5, which achieved only one million tokens in limited scenarios. This advancement in context processing empowers Magic's AI to grasp codebases with greater intricacy, leading to more refined and nuanced suggestions. The company also boasts an impressive technological feat with the creation of the LTM-2-mini model, offering an unprecedented context window of 100 million tokens. This capacity equates to about 10 million lines of code or 750 novels, marking the largest context window available in any commercial model to date. Such capabilities have allowed Magic's models to autonomously execute complex tasks, such as developing a password strength meter for open-source software and creating a calculator with a bespoke UI framework. Despite having a modest team of approximately two dozen engineers and researchers, Magic has drawn substantial financial backing, securing $465 million across various funding rounds. Notable investments include a $320 million infusion led by former Google CEO Eric Schmidt, with participation from other eminent figures and organizations. Earlier rounds saw contributions from industry heavyweights such as CapitalG, Elad Gil, and Sequoia. Magic's partnership with Google Cloud to develop two supercomputers, Magic-G4 and Magic-G5, further underscores its commitment to push the boundaries of AGI research. Utilizing NVIDIA’s H100 and Blackwell GPUs, the collaboration aims to achieve a staggering 160 exaflops of computational power, enabling Magic to enhance its AI capabilities even further. While Magic’s products are not yet available on the market, its audacious research endeavors, robust funding, and strategic collaborations make it a formidable player in the space of generative AI. The company's long-term vision transcends mere code generation, aiming at developing AGI that can reliably solve complex problems, advancing the field in ways previously unimagined.
Featured models
| Model | Released | Context | Input price ($/1M) | Output price ($/1M) | License | Openness |
|---|---|---|---|---|---|---|
| LTM-2-mini | 2024-08-29 | 100m | - | - | Proprietary | Proprietary |
| LTM-1 | 2023-06-06 | 5m | - | - | Proprietary | Proprietary |
Model families
Recent releases
- LTM-2-mini- 2024-08-29
- LTM-1- 2023-06-06
FAQ
Who founded Magic and when?
Magic was founded in 2022 and is associated with San Francisco, California, United States.
What models has Magic released?
Magic ships 2 models across 1 family: LTM.
Is Magic's technology open source?
Magic's models are proprietary.
Where is Magic headquartered?
Magic is headquartered in San Francisco, California, United States.
What is Magic known for?
100M-token context window innovation. Its most prominent tracked family is LTM.
How can I access Magic's models?
Magic's provider availability is tracked on model pages as API and hosting data is verified.
Explore related pages
Last reviewed: 2026-05-19. Data sourced from public lab announcements and provider documentation.