AI Summary
5 min readReflection Unveils Beam, Mistral Drops Le Chonk, and ChatGPT Fakes Cartoonists' Signatures
The biggest AI news today is the arrival of a serious US-based open weights contender. New York City startup Reflection AI has unveiled Beam, a 500-billion-parameter open model that it says rivals China's GLM 5.2 on reasoning while using three to four times less compute, and approaches Alibaba's Qwen 3.8 Max on agentic tasks. The company emerged from stealth last year backed by Nvidia and other investors, and has reportedly held talks to raise a $2.5 billion round at a $25 billion valuation.
How Beam Was Built
Reflection developed Beam through a multi-phase training process that prioritized efficiency. The company started with a small prototype, then scaled up through successively larger algorithms, eventually producing "Beam Base" — the foundation model. That base was trained on a cluster of 6,144 graphics cards using 23.8 trillion tokens from public web sources and commercial datasets, including substantial software code. The company created custom filters for each programming language to remove low-quality files. Remarkably, Beam Base was developed in under four weeks.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Tech Brew Ride Home
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 Le Chonk — Timestamped Outline
- 2 (00:04) **Episode Opening & Headlines** - Brian McCullough introduces the day's top stories: Reflection's open weights model Beam, subscription value comparisons, Meta and Microsoft cutting Claude usage, Mistral's Le Chonk release, and ChatGPT faking New Yorker signatures
- 3 (01:13) **Reflection Unveils Beam: A US Open Weights Model** - New York City-based Reflection launches Beam, an open model designed to rival Chinese competitors on reasoning while using less compute
- 4 (04:53) **Claude vs. ChatGPT: Which Subscription Offers More Value?** - Semi-analysis compares subscription plans for agentic workloads, finding Anthropic's plans offer roughly 5x more API equivalent value than OpenAI's
- 5 (11:18) **Meta and Microsoft Cut Back on Claude Usage** - Both tech giants are reducing internal spending on Anthropic's tools, preferring homegrown alternatives
- 6 (13:58) **Mistral Releases Le Chonk: A 1 Trillion Parameter Open Weights Model** - Mistral launches Mistral Large 4 (Le Chonk), trained on 4,000 Grace Blackwell GPUs at a fraction of competitors' cost
- 7 (16:23) **ChatGPT Faking New Yorker Cartoonists' Signatures** - ChatGPT generates images in New Yorker style and adds real cartoonists' pen names, including Brendan Loper's and Emily Flake's signatures
+ Full timestamped outline available in the app
Show Notes
Nvidia-backed Reflection unveiled Beam, a US open-weights model taking on Chinese labs. Plus, SemiAnalysis said Claude subscriptions were a better deal than OpenAI's, Meta and Microsoft cut back on Claude, Mistral dropped le Chonk, and ChatGPT faked New Yorker cartoonists' signatures.
- NYC-based Reflection unveils Beam, an open model it says rivals GLM-5.2 on reasoning while using 3x-4x less compute and approaches Qwen3.8-Max on agentic tasks (Semafor)
- Nvidia-backed Reflection pitches Beam to businesses and developers as a cheaper option for coding and agent tasks; the startup has held talks to raise $2.5B at a $25B valuation (Bloomberg)
- Analysis: for agentic workloads, Anthropic's plans with Claude Opus 5.5 offer ~5x more API-equivalent value per month than OpenAI's plans with GPT-6.1 Sol (SemiAnalysis)
- Sources: Meta and Microsoft are working to cut internal Claude use; Meta employees using Claude Code dropped to ~30,000 from ~60,000 earlier in 2026 (The Information)
- Mistral releases Mistral Large 4, dubbed “le Chonk”, a 1T-parameter open-weight model for general agentic capabilities, trained on 4,000 Grace Blackwell GPUs (The Deep View)
- ChatGPT is not only using New Yorker cartoonists' style but also adding their signatures to the fake cartoons it creates, which have gone viral on social media (Nieman Lab)
More from this podcast
Tech Brew Ride Home →