The LLMCheck index estimates K2-Horizon-3.7B at ~133 tok/s on an M5 Max at Q4_K_M, a figure from its published bandwidth model, not a measured run. It needs about 4 GB of unified memory. It is MBZUAI IFM's 5.1B dense model, released under the Apache 2.0 licence.
Want to run K2-Horizon-3.7B 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. The bandwidth formula’s typical error for dense models is ±9%, against 75 measured runs of 20 models (how it is computed). Own this Mac? Submit a real run →
| Metric | Value |
|---|---|
| Tokens per second | estimated~133 tok/s |
| Provenance | estimate · how estimates are made |
| Quantization | Q4_K_M |
| Engine | llama.cpp |
| Figure dated | 2026-11 |
| Minimum RAM | 4 GB |
| Parameters | 5.1B |
| 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 (4 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 |
|---|---|---|---|---|---|---|
| estimated133 tok/s | Q4_K_M | llama.cpp | 48 GB | — | LLMCheck estimator | — |
Ollama — the index has not verified an Ollama tag for K2-Horizon-3.7B, 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 → · Hugging Face model card →
Prefer a GUI? LM Studio downloads GGUF and MLX builds in one click. Step by step: install Ollama on a Mac.
MBZUAI's Institute of Foundation Models released this fully open Apache 2.0 dense model on 1 Sep 2026 (3.7B core parameters, 5.1B with its 250K-token embeddings and output head; 512K context; training data and code public). Artificial Analysis ranks it the strongest open model of 4B or fewer parameters, ahead of MiniCPM5-2B. Vendor card: SWE-bench Verified 65.6, Terminal-Bench 2.1 39.7, GPQA Diamond 68.9. Q4_K_M is 3.2 GB; mainline llama.cpp supports it from b11454, while mlx-lm has no k2_horizon architecture yet (#1948), so MLX builds ship their own model file.
Apache 2.0 licence · 5.1B · 512K context · LLMCheck capability 22/50 · see its leaderboard rank →
Estimated, not measured: the index’s minimum for K2-Horizon-3.7B (4 GB — the weights at the quant that fits, plus ~2 GB) plus the KV cache for that many tokens, computed from its config.json. GQA on all 36 layers (8 kv heads x 128). A 16-bit KV cache is the default in llama.cpp, Ollama, LM Studio and mlx-lm; a q8 cache halves the KV part. Coding agents hold 64K tokens or more. Runtimes also add a buffer that grows with context, so leave a few GB spare. macOS lets the GPU use only part of a Mac’s memory by default (figures for an M5 Max Mac, which runs macOS 26 or later): “fits” means within the GPU memory a Mac of that tier can use with the limit raised, a tier tagged raise GPU limit fits only after sudo sysctl iogpu.wired_limit_mb raises it (the tooltip has the exact command), and the last column is the smallest Mac memory size whose default limit holds it (the GPU memory limit).
K2-Horizon-3.7B needs about 4 GB of unified memory. These Apple Silicon Macs have enough to run it.
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 K2-Horizon-3.7B stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard