Reflection AI
Open-source frontier AI lab building mathematical superintelligence to challenge DeepSeek and closed-source labs.
- Founded
- 2024
- Headquarters
- Brooklyn, NY
- Latest Round
- Series B
- Est. Valuation
- ~$8,000M
Investment Thesis
Reflection AI is an open-source frontier AI lab founded in March 2024 by two former Google DeepMind researchers: Misha Laskin, who led reward modeling for DeepMind's Gemini project, and Ioannis Antonoglou, who co-created AlphaGo. The company is positioning itself as both an open-source alternative to closed frontier labs like OpenAI and Anthropic, and a Western equivalent to Chinese AI firms like DeepSeek.
Reflection AI came out of stealth in March 2025 with $130M in total funding combining a $25M seed and $105M Series A. In October 2025, the company raised a massive $2B round at an $8B valuation — one of the largest ever for an early-stage AI company. The round was oversubscribed from an initial $1B target, with investors including Nvidia, Sequoia, CRV, Lightspeed, GIC, DST, and Eric Schmidt. The company has offices in Brooklyn, San Francisco, and London.
The frontier AI landscape is dominated by a handful of closed labs, creating dangerous concentration of power and limiting researchers and developers who want transparency, customizability, and control over their AI stack.
Reflection AI is building open-weight frontier models that compete with GPT-4-class systems, betting that the open-source distribution advantage and community innovation will win the long game — just as Linux did against proprietary Unix.
Open-source models (Llama, Mistral, FLUX) have proven they can match or exceed closed alternatives within months of release, and enterprise adoption of open models is accelerating as companies seek to avoid vendor lock-in.
Training costs are plummeting with better data curation, synthetic data, and architecture improvements, making it newly viable for lean teams to train frontier-quality models without Big Tech budgets.