Beam 1.0 vs SWE-2: Specs & Benchmark Comparison
Beam 1.0 is developed by Reflection AI, while SWE-2 comes from Cognition. SWE-2 was released in September 2026, and Beam 1.0 followed a month later in October 2026. Beam 1.0 has a published size of about 501 billion parameters; Cognition has not disclosed the parameter count of SWE-2.
The two models share 2 published benchmarks. SWE-2 leads on 2 of them. The widest gaps are on DeepSWE 1.1, where SWE-2 scores 73.0% against 43.3%; Terminal-Bench 2.1, where SWE-2 scores 92.8% against 80.1%. Averaged across everything we track, Beam 1.0 sits at 61.6% and SWE-2 at 60.8%.
Beam 1.0 is available through an API at $0.40 per million input tokens with a 1M-token context window. We do not track a hosted API for SWE-2, so pricing and context cannot be compared directly.
| Characteristic | Beam 1.0 | SWE-2 |
|---|---|---|
| Company | Reflection AI | Cognition |
| Release Date | October 16, 2026 | September 10, 2026 |
| Parameters | 501B | — |
| Multimodal | No | No |
| Context (input) | 1.0M | — |
| Context (output) | 131K | — |
| Input Price / 1M | $0.40 | — |
| Output Price / 1M | $1.20 | — |
| Average Score | 61.6% | 60.8% |
| Benchmarks | ||
| DeepSWE 1.1 | 43.3% | 73.0% |
| Terminal-Bench 2.1 | 80.1% | 92.8% |
Visual Benchmark Comparison
Verdict
Beam 1.0 leads in 1 out of 2 comparison categories.
Both models show comparable average scores: Beam 1.0 — 0.6, SWE-2 — 0.6.
Beam 1.0 is newer: released 10/16/2026 vs 9/10/2026.
More About These Models
Related Comparisons
Frequently Asked Questions
Which is better for coding — Beam 1.0 or SWE-2?
Which model is cheaper — Beam 1.0 or SWE-2?
Which has a larger context window — Beam 1.0 or SWE-2?
The Beam 1.0 and SWE-2 comparison is updated for 2026. Data includes benchmark results, API pricing, context window size and other specifications. For more detailed information, visit the Beam 1.0 or SWE-2 page. See also the complete list of AI model comparisons.