The LLMCheck index estimates Kolibri-1 at ~75 tok/s on an M5 Max at Q4_K_M, a figure from its published bandwidth model, not a measured run. It needs about 50 GB of unified memory. It is Aleph Alpha's 78.1B mixture-of-experts model (3.46B active), released under the Apache 2.0 licence.
Want to run Kolibri-1 faster — or step up to bigger models? Find your Mac →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~75 tok/s |
| Provenance | estimate · how estimates are made |
| Quantization | Q4_K_M |
| Engine | MLX |
| Figure dated | 2026-10 |
| Minimum RAM | 50 GB |
| Parameters | 78.1B MoE (3.46B active) |
| Chip bandwidth | 614 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 (50 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 |
|---|---|---|---|---|---|---|
| estimated75 tok/s | Q4_K_M | MLX | 64 GB | — | LLMCheck estimator | — |
Ollama — the index has not verified an Ollama tag for Kolibri-1, 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 →
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 |
|---|---|---|---|---|
| M4 Max | estimated69 tok/s | Q4_K_M | MLX | 546 GB/s |
| M1 Max | estimated54 tok/s | Q4_K_M | — | 400 GB/s |
| M5 Pro | estimated43 tok/s | Q4_K_M | MLX | 307 GB/s |
| M4 Pro | estimated39 tok/s | Q4_K_M | llama.cpp | 273 GB/s |
Aleph Alpha's Apache 2.0 German/English MoE (3 Oct 2026): 78B total, 3.5B active, 256K native context (1M only by extrapolation), with reasoning modes and tool calling. Q4_K_M is ~47 GB, so it needs a 64 GB Mac, and a 3-bit MLX build (~35 GB) fits 48 GB; use llama.cpp b11709+ with GGUFs made by the new converter, or mlx-vlm 0.7.8. A community MLX run measured 52–56 tok/s on an M1 Max with a bundled loader.
Apache 2.0 licence · 78.1B MoE · 262K context · LLMCheck capability 36/50 · see its leaderboard rank →
Kolibri-1 needs about 50 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 Kolibri-1 stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard