Company
AI lab building open-weight frontier language models, backed by Nvidia and other investors; its models include the 501-billion-parameter Beam.
Reflection AI's Beam coding model computes 23B parameters per token but keeps all 501B in memory, according to AWS and Hugging Face. Self-hosting it starts from a roughly 1 TB BF16 checkpoint and multi-GPU sharding.
Reality
Reflection, Aleph Alpha and Mistral released open-weight models for users worried about Chinese ones. Rest of World's report does not publish scores for the three, so a team weighing a switch has to test on its own work.
Reality
Nvidia is in talks to buy Reflection AI or add to the $800 million it has already invested, the Financial Times reported. Owning the open-weight lab, last valued at $25 billion, would put the chipmaker in the model business alongside the AI labs that buy its GPUs.
Perspective Coverage
11 publishersReality
Reflection AI said it will release the weights for Beam, a 501-billion-parameter open model with 23 billion active, later this month. Red-teaming is unfinished, so every score so far is the company's own.
Reality
Reflection AI says its first model, Beam, performs at the level of Z.ai's GLM-5.2 while using three to four times less compute than similar open models. Until outside testers report, those are the company's figures, and they count most with buyers who cannot use Chinese models.
Perspective Coverage
12 publishersReality
Endeavor Catalyst closed its fifth fund at $320 million, taking assets under management past $850 million. For founders outside the big tech hubs it is real money behind two gates: admission to Endeavor's network and a lead investor willing to set the terms.
Perspective Coverage
3 publishersReality
Reflection AI says Beam, its 501-billion-parameter open-weight model, needs three to four times less compute than comparable open models. Buyers have only the company's figures until the weights ship later in October.
Perspective Coverage
13 publishersReality
Reflection AI's 501-billion-parameter Beam beats GLM 5.2 on SWE-bench Pro and trails it on Terminal-Bench, by the company's own scores. Its claimed compute saving covers only part of a serving bill, so planning around it as a US-built open model has to wait for outside tests.
Reality
Reflection AI, backed by about $800 million from Nvidia, is building software that lets companies make their own AI models. The pitch aims at the enterprise spend frontier labs now depend on, though the leverage it creates sits mostly with those labs' largest customers.
Reality
Reflection AI is preparing its first open-weight model after signing compute deals worth $150 million a month with SpaceX and over $1 billion with Nebius. Open weights leave the infrastructure and upkeep with the customer, so how the model deploys is the test buyers can run themselves.
Reality
Tareq Amin wants seven tenths of HUMAIN's funding to come from outside its sovereign parent, and the price of money that size is the kind of anchor tenant his "Switzerland for open models" positioning is meant to hedge against.
Reality
Process-group reconfiguration, one-sided RMA windows and a backend-agnostic Flight Recorder land as core distributed features, alongside native linear algebra on Apple Silicon. The recovery numbers you would plan against are yours to measure.
Reality
Michael Nicolls moves over from Starlink to run the AI data centers, where mobile turbines and portable chillers bought speed at the cost of redundancy, and where Anthropic's $1.25bn-a-month lease can be cancelled.
Reality
The continuation fund buying Lightspeed's OpenAI stake will be managed by Lightspeed. Coller Capital's cheque is the only outside check on the transfer price.
Reality
CoreWeave, Nebius and Cerebras all grew fast and lost more money in Q2 2026. The spending curves say capacity is still being contracted ahead of the demand it serves.
Reality