The LLMCheck index estimates Beam 1.0 at ~27 tok/s on an M5 Ultra at Q4_K_M, a figure from its published bandwidth model, not a measured run. It needs about 199 GB of unified memory. It is Reflection AI's 501.3B mixture-of-experts model (23.4B active), released under the Apache 2.0 licence.
Metal unvalidated: llama.cpp b11701+ loads Beam, but no real-weight run on Apple Silicon has been published; there is no MLX port, and Ollama offers it cloud-only.
This is an LLMCheck index estimate from its published two-term bandwidth model, not a measured run; the index labels it so everywhere it appears. Own this Mac? Submit a real run →
| Metric | Value |
|---|---|
| Tokens per second | estimated~27 tok/s |
| Provenance | estimate · how estimates are made |
| Quantization | Q4_K_M |
| Engine | llama.cpp |
| Figure dated | 2026-10 |
| Minimum RAM | 199 GB |
| Parameters | 501.3B MoE (23.4B active) |
| Chip bandwidth | 1228 GB/s |
Each row as the index holds it, with its source. “Mac RAM” is the memory of the Mac a run used, not the model’s requirement (199 GB). The headline figure follows the reference rule: the lower-median measured plain-decode run at 5 bits or fewer (a run whose quant was not stated counts only when no stated-quant run qualifies), else the estimate; a run marked “not used to anchor estimates” is listed here but never becomes the headline.
| Speed | Quant | Engine | Mac RAM | Context | Source | Note |
|---|---|---|---|---|---|---|
| estimated39 tok/s | IQ3_XXS | llama.cpp | 256 GB | — | LLMCheck estimator | — |
| estimated27 tok/s | Q4_K_M | llama.cpp | 512 GB | — | LLMCheck estimator | — |
Ollama — the index has not verified an Ollama tag for Beam 1.0, so it doesn’t print one (a guessed tag fails with “pull model manifest: file does not exist”). Search the Ollama library →
MLX — pick a build from Hugging Face’s MLX listings, then run it with its repo id:
Downloads: Hugging Face repo → · model card →
Prefer a GUI? LM Studio downloads GGUF and MLX builds in one click. Step by step: install Ollama on a Mac.
| Chip | Speed | Quant | Engine | Bandwidth |
|---|---|---|---|---|
| M3 Ultra | estimated19 tok/s | Q4_K_M | llama.cpp | 819 GB/s |
Reflection AI's first open model (Apache 2.0, weights 16 Oct 2026): a 501B text-only MoE with 23.4B active and 1M context; the card's SWE-bench Verified 81.6 does not name its harness. llama.cpp b11701+ loads it, but real weights have only been validated on CUDA, there is no MLX port, and Ollama and LM Studio offer it cloud-only. The ~3-bit GGUF (197 GB) fits a 256 GB Mac Studio with the GPU memory limit raised and 4-bit (307 GB) needs 512 GB; no Mac measurement has been published.
Apache 2.0 licence · 501.3B MoE · 1000K context · LLMCheck capability 47/50 · see its leaderboard rank →
Beam 1.0 needs about 199 GB of unified memory. These current Apple Silicon Macs have the headroom to run it comfortably:
As an Amazon Associate, LLMCheck earns from qualifying purchases. These affiliate links cost you nothing extra and help keep the LLMCheck index free.
See how Beam 1.0 stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard