Reflection AI

What people say

Reflection AI has no meaningful product-review footprint — no G2 listing, no app-store rating — because reflection.ai lists nothing generally available: the Asimov code agent it launched with no longer appears on the site, which now presents a full stack around open models for developers, enterprises and the public sector. The signal that exists is financial and reputational: founders Misha Laskin and Ioannis Antonoglou (an AlphaGo co-creator) out of Google DeepMind, and a headline $2B raise at an $8B valuation in October 2025, with investors reported to include Nvidia and Sequoia, later reporting pushing the valuation higher.

The dominant sentiment among analysts is skepticism about that valuation-to-shipping gap. Multiple write-ups frame it bluntly — "the $20B startup that has yet to ship" — noting that as of early 2026 Asimov is still waitlisted, the open model isn't public, and the site has product docs but no research papers. Coverage describes a "classic S-curve trap": the market is pricing steep future growth, and the company faces a hard credibility test to justify each round with real product milestones.

Honest read: a heavily-funded team with elite pedigree and a plausible thesis (open frontier models, after an earlier code-comprehension pitch), but essentially unproven by users — buy the team and the story, because there's little shipped product to judge.

The numbers kept climbing in 2026: CNBC reported an April round at a $25 billion pre-money valuation, and the company has signed compute deals with SpaceX, worth up to $6.3 billion, and Nebius, worth $1 billion, according to press reports it links. Its named partners include Dell, the US Department of Energy's Genesis Mission and Shinsegae in South Korea. Reflection promises open weights and research papers, but no model had appeared on Hugging Face by mid-September.

Summary of public user & expert reviews, compiled by RECATOOLS.