OpenAI vs 
Reflection
Verdict
- OpenAI's flagship leads: GPT-6 Astra at 69.7 vs Beam 1.0 at 56.4.
- OpenAI has the better model in 4 of 4 price bands.
- OpenAI's fastest ranked model (GPT-5.4 Mini, 316 tok/s) beats Reflection's (Beam 1.0, 303 tok/s).
- OpenAI publishes more open-weights models (3 vs 2).
Best model per price band
Each provider’s highest-index model below the price ceiling. Purple marks the winner of the band.
| Price band | OpenAI | Reflection |
|---|---|---|
| Under $0.50/M | GPT-5.6 Luna xhigh 59.5$0.450/M | – |
| Under $2/M | GPT-5.6 Luna xhigh 59.5$0.450/M | Beam 1.0 56.4$1.05/M |
| Under $10/M | GPT-6.1 Sol max 69.1$4.00/M | Beam 1.0 56.4$1.05/M |
| Any price | GPT-6 Astra max 69.7$20.00/M | Beam 1.0 56.4$1.05/M |
Line-up
| Metric | OpenAI | Reflection |
|---|---|---|
| Flagship index | 69.7 | 56.4 |
| Ranked models | 45 | 1 |
| Open-weights models | 3 | 2 |
| Cheapest ranked, $/M | $0.128 | $1.05 |
| Fastest ranked, tok/s | 316 | 303 |
| Release cadence, days | 26 | – |
| Last release | 29 Sept 2026 | – |
| Next expected | 14 Oct 2026 – 4 Nov 2026 | – |
Data as of 18 Oct 2026. See OpenAI and Reflection in full.