Reflection
Reflection ranks #16 of 92 providers on BenchLeader by its best model, with 1 ranked model. Its best model is Beam 1.0 at 56.4 on the BenchLeader Index, strongest in long context. Beam 1.0 is priced at $1.05 per million tokens blended. The fastest is Beam 1.0 at 303 tokens per second. All models ship open weights.
- Provider rank
- #16
- by best model
- Best index
- 56.4
- Beam 1.0
- Ranked models
- 1
- 2 open weights
- Cheapest ranked
- $1.05/M
- Beam 1.0
Release timeline
Each model’s best configuration by release date. The top eight are labelled; hover any point for its name.
Not enough dated, ranked models to draw a timeline.
Next release
Not enough dated releases to estimate a cadence.
Fastest ranked model: Beam 1.0 at 303 tok/s.
Head to head
Models
Every Reflection model with a result, one row per model; switch to all configurations to see each reasoning effort.
2 of 2
| # | ||||||||
|---|---|---|---|---|---|---|---|---|
| 1 | New | 56.4±9 | $1.05 | 303 | 7.51 s | 1.0M | 2026-10-16 | 2026-10-16 |
| 2 | New | – | $1.50 | 84 | 0.99 s | 1.0M | 2026-10-05 | 2026-10-16 |
Latest news
- Reflection's Beam aims to undercut Chinese open models with fewer tokens, not cheaper tokensVentureBeat · 16 Oct 2026
- How two DeepMind alumni built a £19bn threat to OpenAIThe Times · 16 Oct 2026
- Nvidia AI Deal Talks Just Gave Bulls a Fresh Reason to Buy NVDA StockBarchart.com · 16 Oct 2026
- Reflection AI open-sources Beam’s weights, post-training recipesSiliconANGLE · 16 Oct 2026
- Nvidia in talks to invest further in Reflection AI or buy it, FT reportsReuters · 10 Oct 2026
- A New Open-Weight Challenger to Anthropic, Reflection, EmergesThe New York Times · 6 Oct 2026
- Nvidia Weighs Buying Reflection AI, the $25 Billion Open-Weight Model Startup It Already BacksTIKR.com · 11 Oct 2026
- Reflection debuts Beam, an open-weight AI model to rival Chinese models at lower compute costTechCrunch · 5 Oct 2026