Hy-Memory-30B-A3B on local hardware
Estimated comfort for repository coding with Aider on a MacBook Pro M4 Pro 48GB.
Hy-Memory-30B-A3BOpen source
tencent · apache-2.0 · 2,145 downloads
Mixture of expertsReasoning / thinking mode
- Parameters
- 30B exact
- Active per token
- ≈1.2B (8 of 128 experts)
- Context window
- 256K tokens
- Attention
- Standard (full attention)
- Layers · KV heads · head dim
- 48 · 4 · 128
- KV cache
- 96 KB/token · 3.0 GB at 32K
- Tool calling
- good (from chat template)
- Thinking
- good (from chat template)
- Vision
- basic (inferred — no supporting signal)
- Commercial use
- allowedRetain the required license and attribution notices
- Import confidence
- medium · capabilities estimated
Weights by quantization
Q4_K_M
17 GB
gguf file size
Q5_K_M
20 GB
gguf file size
Q6_K
23 GB
gguf file size
Q8_0
30 GB
gguf file size
Q3_K_M
13 GB
gguf file size
F16 / BF16
56 GB
gguf file size
- • vision: basic, inferred from missing signals (low confidence).
- • Capability ratings are estimated from model size and name — this model hasn't been assessed by hand.
- • GGUF sizes come from community conversions matched by model name; equivalence and runtime support are not verified. MLX conversion availability is unverified.
Hy-Memory-30B-A3B — 8-bit
Acceptable for your workload
Speed
45–60 tok/s
Excellent
Memory left
2.6 GB
of 48 GB
First word in
≈7 s
Excellent
All the details
- Your setup
- MacBook Pro M4 Pro 48 GB
- Workload
- Repository coding with Aider
- Runtime → tool
- LM Studio → Aider
- Memory used
- 34 GB
- Context
- 32K · up to 32K practical
- Tool support
- Official
- Use-case fit
- Poorly suited
Why?
- LM Studio supports hardware acceleration on this machine (Metal (Apple GPU)).
- Aider can connect to LM Studio (officially supported).
- Generation (45–60 tok/s) should be appropriate for interactive coding.
- Ingesting a prompt of 14K tokens takes ≈7 s; prompt processing keeps up with interactive coding.
- 32K working context is reasonable.
- Only ≈2.6 GB of memory headroom remains for the OS, IDE, browser, terminal and agent.
- Very little free memory: the system may start swapping or kill apps.
- Weak coding capability for interactive coding.
Why “Acceptable” rather than “Comfortable”?
- Workload suitability: Weak coding capability for interactive coding.
- Memory: Estimated peak ≈45 GB of 48 GB; ≈2.6 GB remains for the OS, IDE, browser, terminal and your coding agent to grow into.
- Sustained use: Very little free memory: the system may start swapping or kill apps.
How sure are we?
Medium confidenceEstimated from bandwidth and compute, calibrated against 2 verified benchmarks of other models on this same chip (measured with a different engine, so calibration is partial).
- Calibrated with verified benchmarks of other models on this chip.
- MoE performance varies more between runtimes.
Medium confidence