Reflection AI
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 15
- Huang’s Exit Could Reshape AI Startup Funding
TechCrunch reports that Nvidia invested tens of billions in startups annually and participated in 105 funding rounds this year, often as the largest investor. Investors and founders say Jensen Huang’s departure may reduce capital for AI-native startups, while interim CEO Colette Kress is expected to make investment strategy more conservative; the resulting effects on valuations, dealmaking and competition remain expectations rather than confirmed outcomes.
TechCrunch AI · 🔥 0 - Nvidia’s interim CEO expected to tighten AI financing
The Wall Street Journal, as reported by The Next Web, said interim CEO Colette Kress is expected to tighten Nvidia’s financing of AI cloud providers and neoclouds after Jensen Huang resigned. The company reportedly reduced its commitment to OpenAI’s Ohio data center lease, paused some deals and began requiring more scrutiny of guarantees. Huang had also urged neocloud partners to collect upfront customer payments before renting out Nvidia capacity.
The Next Web · 🔥 0 - Reflection AI Seeks $2.5 Billion After Valuation More Than Triples
The Information reports that Reflection AI set out to raise $2.5 billion early this year, after its valuation more than tripled in six months. Existing backer B Capital wanted to invest again, but the required check was too large for its $500 million main venture fund, so it used a separate pool of capital from its limited partners. The report does not say whether the fundraising was completed.
The Information · 🔥 0 - SpaceX and Tesla form merger committees as AI revenue surges
SpaceX and Tesla have formed separate special committees to discuss a potential merger, Elon Musk said during SpaceX’s quarterly earnings call. Musk called the idea good but said he cannot participate because of a conflict of interest, and the transaction would still require shareholder approval. SpaceX reported Q3 AI revenue of $7.56 billion, which it said was now its largest revenue source; Reuters, as cited by ifanr, said most of it came from renting data-center computing power rather than its own models.
爱范儿 · 🔥 0 - SpaceX Q3 revenue jumps 161% as AI business overtakes Starlink
SpaceX reported Q3 revenue of $14.58 billion, up 161% year over year, and net profit of $4.4 billion versus a $1.1 billion loss a year earlier, according to CNBC and The Next Web. Its Super Intelligence unit generated $7.56 billion in revenue, surpassing Starlink’s $5.25 billion, while AI operating income reached $431 million. SpaceX spent about $19.6 billion on capital expenditure during the quarter, including $16.4 billion for AI and data centers.
The Next Web · 🔥 0 - Bloomberg: Eric Schmidt owns about 1% of Anthropic ahead of expected IPO
Bloomberg reported, citing people familiar with the matter, that former Google CEO Eric Schmidt owns roughly 1% of Anthropic, stemming from tens of millions of dollars he invested in the startup's 2021 Series A round, when PitchBook data put its valuation at $461 million. Investors have discussed an Anthropic IPO valuation of $1.8 trillion to $2 trillion, and Reuters reported some investors are now pushing for around $1.5 trillion after weaker third-quarter results. Anthropic could begin marketing shares as early as the week of November 9, with trading possibly before US Thanksgiving, though the timing and final valuation are unconfirmed; it is unclear whether Schmidt has sold any shares since his investment.
Bloomberg Technology · 🔥 0 - Pentagon's Emil Michael says export controls can rein in AI labs that don't cooperate
Emil Michael, a senior Defense Department official and a vice chair of the White House's Super Intelligence Force, told The Information in a video interview that the executive branch can use tools including export controls to check AI companies that don't cooperate with the government, citing past limits on exporting encryption and enforcement by the Justice Department. He likened frontier labs that call for regulation to a drunk driver asking to be pulled over, and said existing laws are “more than sufficient.” Michael also said the government had already acted: according to his account, Anthropic refused to withdraw a model after the government found it was jailbroken, and within two hours the company was under export controls. He said the Pentagon has labeled Anthropic a supply chain risk, and that there is no discussion about working with the company at the department unless it changes its stance on models having control over humans. He said about 1.7 million of the department's 3.5 million employees now use AI, up from 80,000 a year earlier.
The Information · 🔥 0 - Nvidia bets nearly $40B to become a frontier model lab, not just an AI chipmaker
A PingWest article republished by TechNews says Nvidia has assembled close to $40 billion of deals in under a year — counting a $6 billion non-exclusive technology license plus $1 billion investment in Poolside, a Groq license reported at about $20 billion (undisclosed by the parties), a roughly $12.93 billion pending acquisition of Hugging Face, and early talks with Reflection AI — combining licensing, hiring, equity and acquisitions. It adds that with researchers such as Essential AI founder Ashish Vaswani, who joined in June to work on Nemotron, and its own Nemotron line (Nemotron 3 Ultra, released in June, has 550B total and 55B active parameters and was pretrained on 20 trillion tokens), Nvidia is shifting from AI infrastructure supplier toward a frontier lab that builds its own models and agents, repeating the CUDA-era pattern of betting on the next stage while profits are strong. The article cites last quarter's $96.2 billion revenue, up 106% year over year, and about $59.69 billion in net profit as the financial backdrop.
科技新报 · 🔥 0 - Transformers v5.20.1 adds Reflection AI's Beam coding model
The Transformers v5.20.1 release notes say the update adds Beam, a new model from Reflection AI described as a token-efficient, frontier-scale model focused on coding and agentic use cases. The notes say Beam is built for teams that want to run a powerful model on their own hardware, infrastructure or cloud and customize it. The same patch release fixes a regression that broke model loading on older PyTorch versions and includes fixes for HunYuan-VL image assignment in batched generation and DeepGEMM Mega MoE under expert parallelism.
Transformers Releases · 🔥 0 - Nvidia's nearly $40B AI deal spree signals a frontier-model pivot
An analysis by PingWest says Nvidia has arranged nearly $40 billion in AI deals over the past year involving Poolside, Hugging Face, Groq, Essential AI and Reflection, through licensing, investment, hiring and acquisitions. It reports a $6 billion Poolside technology license plus a $1 billion investment, an announced roughly $12.93 billion Hugging Face acquisition that is still pending, and a Groq licensing deal that media reports put near $20 billion though the companies did not disclose terms. The article argues Nvidia is moving from an AI infrastructure supplier toward a company that also builds frontier intelligence, while cautioning that these arrangements are not all paid acquisition cash.
品玩 实时要闻 · 🔥 0 - Reflection releases Apache 2.0 weights for Beam, a 501B open coding model
Reflection AI released open weights for Beam, its first large language model, on Hugging Face on October 16 under the Apache 2.0 license, in BF16, FP8 and NVFP4 checkpoints. Beam is a sparse mixture-of-experts model with 501B total parameters and 23B active per token, a context window Reflection reports at 1 million tokens, aimed at coding and agentic tasks. Hugging Face, AWS and Baseten detailed launch-day support through transformers, vLLM, SGLang and managed endpoints, alongside post-training recipes and a drafter model for speculative decoding.
Hugging Face · 🔥 0 - Nvidia in talks to acquire or invest more in open-weight startup Reflection AI
Nvidia is in advanced but early-stage talks to acquire US open-weight model startup Reflection AI or to increase its investment in the company, according to Financial Times sources, who cautioned the negotiations could still fall apart. The structure is undetermined: it could be an acquihire that absorbs the core team and licenses technology to avoid lengthy antitrust review, a new equity round, or a partnership giving Reflection more chips and compute; Bloomberg reported Nvidia is discussing boosting its investment or offering computing power. Reflection AI released its first open-weight model, Beam, on October 5; Nvidia had already invested $800 million and the startup was valued at about $25 billion in its March funding round. Both companies declined to comment.
Techmeme · 🔥 0 - OpenAI ARR Said to Be $20B Below Reported Figures as Gross-vs-Net Math Draws Scrutiny
The Financial Times reported that OpenAI's annualized revenue figure, circulated to investors as part of its current funding round, is about $20 billion below a widely cited $70 billion number, and AI stocks fell Thursday on the news. The gap largely reflects a gross-versus-net distinction: Anthropic's figure includes model sales by partners such as Amazon and Microsoft, which take a cut, while OpenAI had been using a net figure. Bloomberg later reported that OpenAI expects to reach $70 billion in annualized revenue by the end of the year, with the assumption that this is net rather than gross. Separately, new open-weight models — Reflection's Beam, Mistral's Large 4 and Germany's Aleph Alpha Kolibri — launched with an AI-sovereignty pitch, though investors said national origin matters little to most of their portfolio companies.
Newcomer · 🔥 0 - Disruptive targets up to $10B AI-era late-stage fund
Dallas-based investment firm Disruptive is targeting up to $10 billion for a late-stage fund and has secured $7.5 billion, according to sources cited by The Wall Street Journal. The firm, founded in 2012, has backed AI startups including Groq and Reflection AI.
Techmeme · 🔥 0 - Reflection AI launches Beam, a 501B open-weight model
Reflection AI unveiled Beam on October 5, its first open-weight model, aimed at coding, reasoning and agentic workloads. Beam is a sparse mixture-of-experts model with 501 billion total parameters and 23 billion active, pretrained on 23.8 trillion tokens with a 1 million-token context window. The company says it approaches GLM-5.2 on some advanced reasoning tasks using roughly one-third to one-quarter of the inference compute, but the claims have not been independently verified; weights, a technical report and tools are slated for release later in October.
IT之家 AI · 🔥 0
Experience and discussion from the community
Share my Reflection AI experienceAsk about Reflection AI
Nobody has shared their experience with Reflection AI yet.