Observed Signal · Sep 8, 2026 · Technical Release · Source: Trending Topics (DACH/CEE Innovation & Tech) · Impact: 4/5 · Sentiment: Positive
Magic AI Claims Frontier-Level Pretraining for Under $1M
Magic, an AI startup co-founded by Austrians Eric Steinberger and Sebastian De Ro, claims a major breakthrough in pretraining efficiency. Its new recipe reportedly matches DeepSeek V4 Pro's base model quality using 50x less compute, costing about $500,000 on Nvidia GB200 systems, and is over ten times more compute-efficient than leading open-weight models. Scaling to roughly $4 million, Magic says it outperforms all public base models on perplexity evaluations, comparing against DeepSeek V4 Pro, Kimi K2, and Nvidia's Nemotron 3 Ultra, while excluding closed models from Anthropic, Google, and OpenAI. The company has raised over $460 million, with a $320 million round valuing it at $1.5 billion, and partners with Google Cloud for tens of thousands of GB200 chips. Magic has not yet released a model, and all claims are self-reported.
Magic's claim of frontier-level pretraining at a fraction of the cost could disrupt the AI model training landscape, making it relevant for AI infrastructure and potentially affecting the broader tech ecosystem.
Track Magic Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your Polaris7 workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Magic claims its pretraining recipe matches DeepSeek V4 Pro's quality with 50x less compute, costing ~$0.5M on GB200, and is 10x more efficient than leading open models.
- Scaling the recipe to ~$4M, Magic says it beats all public base models on perplexity evaluations.
- Magic has raised over $460M, with a $320M round valuing it at $1.5B, and partners with Google Cloud for GB200 compute.
- Magic has not yet released a model, and its claims have not been independently verified.
- Magic's model achieves a 72% pass rate on math problems in a short RL run, compared to GPT-6 Astra at 100% and Kimi K3 at 75%.
Connected Companies & Entities
11 Entities mapped“Magic is an AI startup founded by Eric Steinberger and Sebastian De Ro, claiming frontier-level pretraining efficiency.”
“Magic's compute costs refer to Nvidia's GB200 systems, and Nvidia's Nemotron 3 Ultra is used as a comparison model.”
“Magic compares its pretraining recipe to DeepSeek V4 Pro and V4 Flash.”
“Moonshot's Kimi K2 and Kimi K3 are used as comparison models.”
“Magic brought in the inference provider Fireworks for an independent check of baseline figures.”
“Magic announced a partnership with Google Cloud to build compute capacity with tens of thousands of GB200 chips.”
“Alphabet participated in a $23 million funding round for Magic.”
“Jane Street backed a $320 million round for Magic.”
“Atlassian backed a $320 million round for Magic.”
“Sequoia backed a $320 million round for Magic.”
“CapitalG backed a $320 million round for Magic.”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
General Intuition in talks to raise $300M at $2B valuation
General Intuition, a New York startup building foundation 'world models' that train embodied AI agents using interactive gameplay video, is in talks to raise about $300 million at a valuation just over $2 billion. The company spun out of Medal eight months earlier after a $134 million seed round and is reportedly backed or pursued by high-profile investors, including Jeff Bezos, Eric Schmidt, Khosla Ventures and General Catalyst. General Intuition trains agents on Medal’s large gameplay dataset (reportedly 2 billion videos per year from 10 million monthly active users) and plans to expand compute capacity to support a product launch by late summer or early fall. The fundraising would accelerate scaling amid growing competition in world models from companies like Runway, Decart and World Labs.
General Intuition Raises $320M to Train AI with Game Data
General Intuition raised $320 million at a $2.3 billion valuation to advance an agentic foundation model trained on hundreds of millions of hours of gameplay and embedded player action labels from Medal. The startup, spun out of Medal, says its single world-model-style agent can generalize from gameplay to simulation and physical embodiment (robots, drones) and demonstrated a quadruped robot fine-tuned with eight minutes of real-world data. The round was led by Khosla Ventures with participation from General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg and researchers from Google DeepMind and MIT. Proceeds will primarily scale compute (deal with CoreWeave) and expand API access by the end of summer; the company also launched Nerve, a jobs marketplace to monetize gamers and build a data/teleoperation talent pipeline.
Yann LeCun's AMI Labs Secures $1.03B for AI Research
AMI Labs, an AI startup co-founded by Turing Award winner Yann LeCun after his departure from Meta, raised $1.03 billion at a $3.5 billion pre-money valuation to develop ‘world models’—AI that learns from real-world data rather than only language. Led by CEO Alexandre LeBrun, AMI positions itself as a research-first lab building on concepts such as JEPA (Joint Embedding Predictive Architecture). The round was co-led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions and included strategic investors and angels (including Tim and Rosemary Berners-Lee, Mark Cuban and Eric Schmidt). The company plans multi-year research work, will publish papers and open-source code, and names Nabla as its first disclosed partner. AMI will prioritize compute and high-quality talent across Paris (headquarters), New York, Montreal and Singapore, and currently does not plan to generate revenue in the near term.
Receive all future signals for Magic & competitors in your briefing
Never miss material M&A deals, leadership changes, account wins, or product rollouts. Polaris7 monitors 20,000+ primary sources 24/7, delivering verified telemetry directly into your executive briefing.