KAT-Coder-V2.5 runs at 74 tok/s (74.2 as recorded) on a 48 GB M4 Max, a community run reported by llm-bench.io (LM Studio, quantization not stated, up to ~5.9k tokens used). It needs about 24 GB of unified memory. It is Kwaipilot's 35B mixture-of-experts model (3B active), released under the Apache 2.0 licence.
Want to run KAT-Coder-V2.5 faster — or step up to bigger models? Find your Mac →This is a community run: a real run, reported by its source and linked below. The index records it as given and never adjusts it.
| Metric | Value |
|---|---|
| Tokens per second | community74 tok/s |
| Provenance | community run · llm-bench.io |
| Quantization | quantization not stated |
| Engine | LM Studio |
| Context | up to ~5.9k tokens used |
| Figure dated | 2026-09-19 |
| Minimum RAM | 24 GB |
| Parameters | 35B MoE (3B active) |
| Chip bandwidth | 546 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 (24 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 |
|---|---|---|---|---|---|---|
| community74.2 tok/s | quantization not stated | LM Studio | 48 GB | up to ~5.9k tokens used | llm-bench.io | 40-core GPU part (16-core CPU), one anonymous llm-bench.io run. LM Studio with MLX-format OptiQ weights by model id; the source's quantization field is empty, so the row says 'unstated'. Thinking on; the figure is the mean of four task scenarios (73.2-75.6 tok/s each, at most 5,907 context tokens), with prefill reported separately. |
| estimated80 tok/s | Q4_K_M | MLX | 48 GB | — | LLMCheck estimator | — |
Ollama — the index has not verified an Ollama tag for KAT-Coder-V2.5, 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: Search the Ollama library → · model card →
The 74 tok/s figure above came from LM Studio at quantization not stated; other runtimes and quants run at different speeds.
Prefer a GUI? LM Studio downloads GGUF and MLX builds in one click. Step by step: install Ollama on a Mac.
Kwaipilot's Apache 2.0 agentic coder; the open-weight 'Dev' release is a 35B MoE with 3B active, built on Qwen3.6-35B-A3B. About 20 GB at 4-bit, so it fits a 24 GB Mac; a community oMLX run measured 89.3 tok/s on an M5 Max (OptiQ 4-bit, MTP off), below the index's earlier 117 estimate.
Apache 2.0 licence · 35B MoE · 256K context · LLMCheck capability 39/50 · see its leaderboard rank →
KAT-Coder-V2.5 needs about 24 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 KAT-Coder-V2.5 stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard