Reflection AI launches Beam 501B open-source model
Reflection AI Inc. has launched Beam, an open-source large language model featuring 501 billion parameters. The release follows the startup reaching a $25 billion valuation and reportedly entering a $6.3 billion agreement with SpaceX to rent Nvidia GB300 NVL72 systems used for the initiative. Claims are as reported; this summary makes no determination about accuracy or significance.
- First detected
- Oct 6, 2026
- Last updated
- Oct 6, 2026
High confidence
Independently reported by multiple publishers.
Corroborated
Corroborated across 2 independent sources
What does this mean?
Corroboration measures how many genuinely independent sources support the event. Confidence measures how reliable the available evidence appears.
No material changes recently.
Stable
No recent reporting has materially changed the known facts.
Follow this development to see meaningful updates as new evidence emerges.
Save keeps this for later. Follow tracks meaningful changes as new evidence emerges — it shapes your Following Feed, alerts, and digest eligibility, and doesn't promise an instant notification.
Why it matters
The debut introduces a massive 501-billion-parameter open-source alternative into the frontier model ecosystem. It also highlights the extreme capital and compute commitments startups are deploying via specialized hardware arrangements to train and serve very large foundational models.
Coverage
Independent reporting2 sources
Primary/vendor sources vs independent reporting
Primary / vendor source: information published directly by the company, organization, government body or project involved. Useful as a primary source, but not independent confirmation.
Independent reporting: reporting or analysis from a source independent of the organization making the underlying claim.
Timeline
Oct 23, 2026
Development status changed
Developing → Stable — new reporting is no longer substantially changing the known facts
Oct 7, 2026
Corroboration updated
The set of supporting sources changed.
Development status changed
New → Developing — new reporting is still materially changing this development
Oct 6, 2026
New reporting added
Reflection's Beam becomes the most capable open-weight model built outside ChinaThe DecoderIndependent reporting
Development detected
New reporting added
Reflection AI debuts open-source Beam model with 501B parametersSiliconANGLEIndependent reporting
Related Intelligence
- ReportAlso involving China
Alice Huang’s Companies Invested in DeepSeek, Promising Nvidia Chips
According to reporting by The Information citing sources with direct knowledge, corporate entities linked to Alice Huang invested in Chinese AI firm DeepSeek with plans to supply export-restricted Nvidia Blackwell processors. The reported arrangements were intended to establish a channel for delivering advanced AI computing hardware that is prohibited under export restrictions to China.
The Information - ReportAlso involving China
Chinese AI sell-off deepens over fears of tougher Nvidia chip restrictions
The Financial Times reports a deepening sell-off in Chinese artificial intelligence assets driven by investor concerns over stricter restrictions on Nvidia hardware. The report notes that former chief executive Jensen Huang had previously supported maintaining chip sales to China.
Financial Times (AI) - ReportAlso involving DeepSeek
DeepSeek Aims to Close $7.5 Billion Funding Round by End-October
The Information reports that AI firm DeepSeek is seeking to complete a $7.5 billion second funding round by the end of October. The fundraising effort comes ahead of a planned public listing on the Shanghai Stock Exchange and is supported by annualized revenue reaching $1 billion.
The Information - ReportAlso involving China
Paris-based Mistral releases "Le Chonk," a strong coding model that still trails China overall
Mistral has released the open weights for Large 4, an open-weight, trillion-parameter model trained entirely in Europe nicknamed "Le Chonk." Mistral claims the model outperforms open competitors at coding tasks, while independent testing indicates it operates quickly and at relatively low cost. However, it trails leading Chinese models, posting a score of 26 compared to 32 from DeepSeek V4.1 Flash and GLM-5.3.
The Decoder