Beam 1.0
Beam 1.0 is a Reflection open-weights reasoning model, released 16 Oct 2026. Its best configuration ranks #182 of 761 on the BenchLeader Index at 56.7 ±8.6, in the upper half. The ± is the point: 384 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in long context (64) and lowest in agents & tools (43). At $1.05 per million tokens blended it is mid-priced. Output speed of 228 tokens per second puts it in the fastest quarter, with a first answer in 10.1 s. Last measured 16 Oct 2026.
- Blended price
- $1.05/M
- $0.700 in · $2.10 out
- Output speed
- 228 tok/s
- measured by Artificial Analysis
- First answer
- 10 s
- Context
- 1.0M
- Full answer
- 12 s
- median, reasoning included
- Cost per run
- $0.566
- one full Intelligence Index run
- Released
- 16 Oct 2026
- Overall index57
- Reasoning63
- Agents & tools43
- Knowledge63
- Long context64
- Composite64
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source |
|---|---|---|
| Humanity's Last Exam (AA)not in index | 34.7%#110 | Artificial Analysis |
| CritPt | 13.1%#93 | Artificial Analysis |
Coding
| Benchmark | not stated | Source |
|---|---|---|
| SciCode (AA)not in index | 48.5%#111 | Artificial Analysis |
| Terminal-Bench 4.0 (AA)not in index | 3.0%#96 | Artificial Analysis |
Agents & tools
| Benchmark | not stated | Source |
|---|---|---|
| GDPval-AA v2.1not in index | 33.9%#102 | Artificial Analysis |
| AutomationBenchnot in index | 48.5%#42 | Artificial Analysis |
| GDP.pdfnot in index | 11.6%#46 | Artificial Analysis |
| AA-Briefcase v1.1not in index | 976#50 | Artificial Analysis |
| CWE-Bench (AA) | 49.2%#15 | Artificial Analysis |
| DeepSec-Bench (AA) | 6.9%#17 | Artificial Analysis |
| CyberGym end-to-end (AA) | 45.0%#8 | Artificial Analysis |
Knowledge
| Benchmark | not stated | Source |
|---|---|---|
| AA-Omniscience | 1.5#108 | Artificial Analysis |
| AA-Omniscience: accuracynot in index | 33.1%#132 | Artificial Analysis |
| AA-Omniscience: non-hallucinationnot in index | 52.8%#103 | Artificial Analysis |
Long context
| Benchmark | not stated | Source |
|---|---|---|
| AA-LCR | 78.0%#109 | Artificial Analysis |
Composite
| Benchmark | not stated | Source |
|---|---|---|
| AA Intelligence Index v4.3.2 | 30.3#107 | Artificial Analysis |
| AA Cyber Indexnot in index | 3369.9%#11 | Artificial Analysis |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0009 | – | 11.4 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0097 | – | 12.7 s |
| Code edit | 6,000 / 1,500 | $0.0074 | – | 16.7 s |
| Agentic coding session | 60,000 / 4,000 | $0.050 | – | 27.7 s |
| Structured extraction | 2,000 / 200 | $0.0018 | – | 11.0 s |
More from Reflection
See also
Data as of 18 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Beam 1.0: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/reflection-beam, data as of 18 Oct 2026.